Soft Self-Healing Nanocomposites

This review article provides an overview of the research and applications of soft self-healing polymers and their nanocomposites. A number of concepts based on physical and chemical interactions have been explored to create dynamic and reversible gel and elastomer networks, each strategy presenting its own advantages and drawbacks. Physical interactions include supramolecular interactions, ionic bonding, hydrophobic interactions and multiple intermolecular interactions. Such networks do not require external stimulus and are capable of multiple self-healing cycles. They are generally characterized by a rapid but limited healing efficiency. The addition of nanofillers enhances the mechanical strength of the soft networks as in conventional gels and elastomers, and do not compromise the network healing dynamics. In certain cases nanofillers moreover trigger the healing process through e.g., multi-complexation processes between each component. Chemical interactions include Diels-Alder reactions and disulphide, imine, boronate ester or acylhydrazones bonding, and are usually triggered with an external stimulus. The resulting healing is efficient, leading to good mechanical properties, but is generally slow at ambient temperatures, and dynamic chemical interactions are only reversible at higher temperatures. Conductive nanofillers were reported to speed up the healing process in such systems owing to their energy absorption properties. The challenges with nanofillers remain their functionalization and dispersion within the self-healing formulations. Soft self-healing gels and nanocomposites find applications in engineering such as coatings, sensors, actuators and soft robotics, and in the bio-medical field, including drug delivery, adhesives, tissue engineering and wound healing.


INTRODUCTION
Material's research has traditionally focused on the design of new materials that are intrinsically resistant to chemical and mechanical damage; however, material properties degrade with time, and during service owing to unavoidable impacts and stresses inducing cracks. Since few decades, nature has inspired scientists to develop new chemistries and processes to restore material properties after mechanical and functional damage has occurred. Small wounds can generally be healed in biological systems (Harrington et al., 2016), but large body parts such as limbs can also be regenerated, such as in the axolotl, a type of Mexican salamander (Kragl et al., 2009). This inspiration has led to the development of new materials, in particular polymers having self-healing properties (Binder, 2013;Herbst et al., 2013;Yang and Urban, 2013;Campanella et al., 2018;Dahlke et al., 2018;Diba et al., 2018;Ding F. et al., 2018;Kim et al., 2018;. Compared to conventional polymers, self-healing polymers demonstrate the ability to reform broken bonds via a physical process or a symbiosis of physical and chemical processes. The design of materials having self-healing abilities requires a multidisciplinary process including mechanics, process dynamics, and chemistry. Several realistic strategies have been formulated over the past two decades (Chen et al., 2002;White et al., 2002;Gould, 2003), including elastomers (Sordo et al., 2015), coatings (Samadzadeh et al., 2010;Bailey et al., 2015;Zhang et al., 2018), and composite materials (Cho et al., 2009;Cohades et al., 2018), which not long ago appeared a creation of vivid imagination. A number of review articles comparing different healing concepts in bulk as well as composite materials is available (Caruso et al., 2009;Blaiszik et al., 2010;Binder, 2013;Herbst et al., 2013;Yang and Urban, 2013;Bekas et al., 2016;Cohades et al., 2018).
In practice, two interactive effects are observed in self-healing systems. First, in order to close the damage cavity a certain physical flow of a "mobile phase" has to occur within the network structure (Yang et al., 2015b). Second, in order to ensure the restoration of the functional and mechanical integrity, the polymer network needs to be rebuilt. One of the earliest self-healing techniques relies on an extrinsic capsule/vascule based concept, whereby a microencapsuled healing agent is incorporated within the polymer. The stress-induced rupture promotes monomer delivery to the cracks, in turn inducing an appropriate polymerization reaction between the components, often involving a catalyst (White et al., 2002;Brown et al., 2005;Wilson et al., 2008;Guadagno et al., 2010;Coope et al., 2011). However, capsules act as voids in the materials, the encapsulated agent is consumed after a first damage event, thus reducing the ability to heal multiple times, and the stability of catalyst or healing agent is often limited (Brown et al., 2004;Blaiszik et al., 2008;Jin et al., 2013).
The present review focuses on the alternative intrinsic self-healing approach, where the polymer itself repairs after damage by the utilization of dynamic/reversible interactions that can be broken and renewed under different physicochemical conditions. Intrinsic self-healing presents several advantages in comparison to extrinsic self-healing due to multiple times of healing at the same location without help of any healing agent and catalyst. Two main strategies, based on physical interactions and chemical interactions have been implemented to create intrinsic self-healing polymers, as sketched in Figure 1. Reversible bonding can be either non-covalent, including hostguest interactions, π-π stacking, hydrogen bonds, ionic and hydrophobic interactions, metal coordination, or covalent, such as imine bonds, disulphide and diselenide bridging, Diels-Alder reaction, reversible boronate ester, and acylhydrazones bonds. As detailed in the following, the reversibility of covalent bonds between different functional groups can be induced by external stimuli, and can be influenced by a change in temperature, pH, catalytic activity, irradiation, moisture, etc. (Garcia, 2014).
Focus of the review is more specifically on soft materials (gels, elastomers) and their nanocomposites, that have gained strong interest in recent years. Application examples include actuators, sensors Chortos et al., 2017;Han et al., 2017d), flexible conductors (Wu and Chen, 2016b), tissue engineering scaffolds, drug delivery systems (Bowden et al., 1998;Geckil et al., 2010;Vig et al., 2013). Conventional soft materials have low stiffness and ultimate strength of a few MPa or tenths of MPa in the case of tough hydrogels, which greatly limits their application. The addition of nanofillers is thus highly advantageous not only for enhancing their mechanical properties by introducing higher stiffness reinforcement (Tee et al., 2012) but also to improve their functional (for example thermal or electrical conduction) and biological (Annabi et al., 2014;Lau and Kiick, 2015) properties. Traditional nanofillers include carbon nanotubes (CNTs) (Gopal et al., 2010;Qi et al., 2013), graphene (Potts et al., 2011;Huang et al., 2012;Shaygan Nia et al., 2015), ceramic oxides (Mera et al., 2013), nano silica (Engel and Kickelbick, 2013), and nanocellulose (Galland et al., 2014). As an example, the tensile strength of polyvinyl alcohol was increased up to 60 times with the addition of cellulose nano-whiskers (Ding Q. et al., 2018). Nanofillers have also been exploited to promote self-healing properties. For instance, conducting nanofillers with high thermal conductivity (graphene, CNT, metals) are used as nanoscale heaters to accelerate the self-healing mechanisms within the polymer matrix Dai et al., 2018). In fact, nanocomposite materials feature very large specific interfacial areas, usually above 100 m 2 /g. Careful design and control of the polymer-nanofiller interfacial bond dynamics (including through functionalization of the particles and/or the polymer), which remain challenging to optimize nanofiller dispersion and process rheology, represents a further option toward intrinsic self-healing.
Keeping in mind the properties and applications of soft materials, this article reviews the strategies for synthesizing selfhealing soft materials based on the concept of constitutional dynamic chemistry (intrinsic self-healing) relying on physical and chemical interactions. For each case, the effect of nanofillers addition as a tool to enhance the self-healing properties of the soft materials is discussed in details. We also illustrate the emerging applications of self-healing soft nanocomposites by highlighting their usage in various fields.

Physical Interactions Based Soft Self-Healing Materials and Their Nanocomposites
This section reviews the strategies for the preparation of physical self-healing gels and elastomers based upon dynamic networks of non-covalent interactions, including supramolecular interactions, ionic bonding, hydrophobic interactions, multiple intermolecular interactions, and nanocomposite interactions. An overview of the reaction types, which are discussed herein, is provided in Table 1. Please note that it is difficult to compare the various materials as the methods to assess their mechanical properties and healing efficiency are different. Relying on  (Binder and Zirbs, 2007;de Greef and Meijer, 2008). For example, poly(ethylene glycol) based hydrogels were synthesized while functionalizing the prepolymer with quadruple hydrogen bond (UPy groups); due to their hydrophilic nature PEG-moieties can accommodate water, whereas the UPy groups give strength and elasticity while assembling into hydrogen-bonded arrays (Van Gemert et al., 2012). A heart shaped object demonstrated a shape persistent and self-healing behavior when cut into two pieces, which was mended by simply pressing both halves together (Figure 2).
In another example, an ionic self-healing mechanism involves charged polymer chains cross-linked with oppositely charged ions or polymer chains. Relative to organogels, ion gels exhibit high chemical stability and ionic conductivity, and are thus advantageous for a wide range of applications from soft actuators (Imaizumi et al., 2012) to electrolytes for lithium ion batteries (Kitazawa et al., 2018). A self-healing and adhesive ionic gel was prepared by the mixing of polycation poly(allylamine hydrochloride) (PAH) with pyrophosphate (PPi) and tripolyphosphate (TPP) (Huang et al., 2014). Due to highly dense ionic crosslinking, a high storage modulus was observed (G' ∞ ≈ 4 × 10 5 Pa). Due to stabilization and functional groups rearrangement, the self-healing performance of physically crosslinked hydrogels decreases when increasing the time to place back together the cut surfaces (Herbst et al., 2013). A time independent single-component zwitterionic hydrogel with a capability for spontaneous healing under physiological conditions has been reported (Bai et al., 2014), where carboxylbetaine acrylamide (AAZ) was physically crosslinked via electrostatic interactions among zwitterionic moieties, resulting in a time-independent healing behavior. In addition, their self-healing process does not require any external stimulus; however, a decrease in self-healing was observed with increase in crosslinking density (Bai et al., 2014).
Electrostatic interaction based physically crosslinked zwitterionic hydrogels were developed using carboxybetaine acrylamide, 7-acrylamidoheptanoate acid and hydroxyhexylacrylamide (Bai et al., 2014). Interestingly the reported hydrogel demonstrates a time-independent healing behavior, in that the separation time of cut fragments does not compromise healing efficiency. Wei et al. reported electrostatic interactions based self-healable supramolecular luminescent hydrogels from ABA triblock copolymers [Poly(2-(2-guanidinoethoxy) ethyl methacrylate)-b-poly(ethylene oxide)-b-poly(2-(2-guanidinoethoxy) ethyl methacrylate)] and polyoxometalates (Wei et al., 2014), where the mechanical strength of the hydrogel can be simply tuned by altering the ionic strength, concentrations, and copolymer composition. A reversible electrostatic interaction promoted ionic selfhealing mechanism was emphasized by Varghese et al., where a charged polymer chain crosslinks with an oppositely charged polymer chain or with oppositely charged ions. However, based on experimental conditions the materials also illustrate the hindrance in self-healing due to electrostatic repulsion when the polymeric ions were deprotonated (Phadke et al., 2012). Recent years have observed a growing attention in the development of nature inspired dopa-modified polymeric   structures, in particular to explore the biomimetic adhesives for tissue engineering (Barrett et al., 2013;Kim et al., 2014). Catechol induced polyelectrolyte coacervate bioadhesive has been developed using a poly(acrylic acid) backbone decorated with catechol appendants in presence of weak divalent cross-linker Zn 2+ , where negative charged carboxylic groups from poly(acrylic acid) electrostatically interact with zinc chelated mono-catechol group, thus leading to a quick generation of the dense coacervates. A double-network (DN) based self-healing material with improved mechanical properties has been reported, where a symbiosis of covalent and non-covalent crosslinking network had been employed (Gong, 2010). However, as the chemically covalent-cross-linked network cannot reversibly recover to its original state, the DN soft materials exhibit poor self-healing efficiency . Thus, a multiple noncovalent cross-linked network could be more advantageous for designing self-healing materials with high mechanical properties. A double network based poly(N-acryloyl glycinamide-co-Nbenzyl acrylamide) containing a triple amide in one side group was reported by Feng et al. The triple amide based soft material demonstrated a good shape memory ability, high mechanical strength (1.1 MPa) and about 95% self-healing efficiency at room temperature (Feng et al., 2018).
Most of the widely explored soft materials are also now often combined with nanosized inorganic components to form nanocomposites, as summarized in Table 2. The addition of stiff nanofillers is strategic not only for enhancing the mechanical properties of the soft matrix (Tee et al., 2012) but also to improve their self-healing properties Dai et al., 2018).  Zhong et al. (2015a) proposed a tough and highly stretchable self-healing hydrogel based on silica nanoparticles, ferric ions and poly(acrylic acid) (PAA). The hydrogel demonstrated excellent toughness (tensile strength 860 kPa, elongation at break ∼2,300%). It was suggested that the stretchability and toughness arose from the reversible cross-linking interactions between polymer chains, helpful for energy dissipation through stress triggered dynamic process. Exfoliated sodium montmorillonite based self-healing hydrogel was formed by in situ polymerization of acrylamide monomers (Gao et al., 2015). These hydrogels demonstrated good stretchability with a high fracture toughness (10.1 MJ m −3 ) and fracture strain up to 11,800%, however, exhibited slow recovery of mechanical properties (RT, 5 days) and required multiple dry/re-swell handlings. A host-guest interaction based healable, stable, and recoverable nanocomposite hydrogel was developed using a three-component architecture consisting of poly(vinyl alcohol) (PVA), cucurbit[8]uril (CB) and cellulose nanocrystals (McKee et al., 2014). These nanocomposite hydrogels exhibit significant mechanical strength and stiffness with a storage modulus of 10 MPa. Immediate self-recovery was shown via step-strain measurements at RT and rapid self-healing abilities were  (Ding Q. et al., 2018). Reproduced with the permission of 2018 American Chemical Society.
Frontiers in Materials | www.frontiersin.org observed (within 30 s, RT). Also the hydrogel was able to rapidly self-heal (within 10 s) once re-joined even after 4 months.
In another example, self-recovering hydrogels have been formed based on a viscoelastic polyvinyl alcohol (PVA)-borax (PB) gel matrix and nanostructured CNFs-PPy (cellulose nanofibrils-polypyrrole) complexes that synergizes the biotemplate role of CNFs and the conductive nature of PPy. The CNF-PPy complexes not only tangle with PVA chains though hydrogen bonds, but also form reversibly cross-linked complexes with borate ions (Ding Q. et al., 2018). The multi-complexation between each component leads to the formation of a hierarchical three-dimensional network (Figure 3). Owing to a combined reinforcing and conductive network inside the hydrogel (2 wt.% CNFs-PPy), it demonstrated high storage modulus (∼0.1 MPa) and nominal compression stress (∼22 MPa), about 60 and 2,240 times higher than those of pure PVA hydrogel, respectively.
Mussel inspired, metal coordination bonds based on reversible dynamics have been demonstrated to function as dynamic cohesive cross-linkers in bulk polymer materials (Amstad et al., 2009). Niels-Andersen et al. incorporated iron oxide nanoparticles (Fe 3 O 4 NP) in catechol modified polymer networks to obtain hydrogel cross-linked via reversible metalcoordination bonds at Fe 3 O 4 NP surface (Li et al., 2016b). The structure provides solid-like yet reversible hydrogel mechanics. Zirconium hydroxide has been utilized as nanofiller in a poly(acrylamide) to produce hydrogen bonding interaction based self-healing gel. The gels exhibited good mechanical properties with 114 KPa-404 KPa tensile strength and 284-723% elongation at break . A metal-ligand network based highly stretchable self-healing elastomer was developed consisting of 2,6-pyridinedicarboxamide ligands that coordinate to Fe(III) centers through weak carboxamido-iron ones, a strong iron-pyridyl one. The material demonstrated self-healing at a temperature as low as −20 • C (Figure 4) (Li et al., 2016a).
Carbon nanomaterials (CNT, graphene), which show good compatibility with polymers due to their large π-conjugated system, have been widely used to produce mechanically enhanced composite materials. Such nanofillers moreover improve the chemical stability, the electrical and thermal conductivity (Balandin et al., 2008), and the microwave and infrared (IR) absorbing capacity of polymers (Sui et al., 2011). CNT based supramolecular hydrogels directed by hierarchical hydrogen bond (a mixture of strong and weak bond) interaction has been reported (Du et al., 2014b), where the gel was obtained by mixing multiwall carbon nanotubes (MWCNT) with polyethylene polyamine (PAA). The    properties of the gel can be easily tailored by controlling the ratio of strong and weak hydrogen bonds. The autonomous healing gel demonstrated a multiple responsiveness (NIR light, pH, and thermal) including temperature dependent reversible adhesion behavior. A host-guest interaction based selfhealable conducting elastomer consisting of singe-walled carbon nanotubes (SWCNT) and poly(2-hydroxyethyl methacrylate) (PHEMA) was reported (Guo et al., 2015). The elastomers sustain a high conductivity up to 180% stretching, however, under large strain the conductivity decreases. This might be due to partial breakdown the SWCNT network that leads to a decrease in the total number of conduction paths. Wang et al. have developed a rapid room temperature self-healing elastic nanocomposite material by combining the hydrogenbonded polymer and graphene oxide as a cross-linker , where hydrogen bonding network in polymer    chain provides self-healing capability, while GO enabled good mechanical properties. A symbiosis of shape memory and self-healing was also achieved for graphene oxide based polyurethane (GO-PU) nanocomposites , where fracture of the composite was effectively healed by exposure to direct sunlight and MW. The nanocomposites were effectively healed within 30-50 s under low MW power (360 W) and within 5-7 min under direct sunlight. During the healing process, GO absorbed energy from the stimulus, and then transferred this energy to the HPU matrix, where the soft segment of the HPU melted (low T m ∼50 • C). Thus, the crack could repair with a higher mobility of the soft segment of HPU. To achieve combined attributes of shape memory, self-healing and self-cleaning properties, hyperbranched polyurethane (HPU)-TiO 2 /reduced graphene oxide (TiO 2 /RGO) nanocomposites were reported , where the fabricated nanocomposite exhibited composition and dose-dependent mechanical properties with excellent shape recovery ratio (91-95%) as well as shape recovery rate (1-3 min) under exposure to sunlight. The presence of a high amount of RGO (0.5-1 wt.%) in the nanocomposite helps in rapid and efficient healing, whereas a high amount of TiO 2 nanoparticles (5-10 wt.%) aids in achieving good selfcleaning properties.

Chemical Interactions Based Self-Healing Soft Materials and Their Nanocomposites
This section covers the strategies for the preparation of chemical self-healing soft materials based upon dynamic networks of covalent interactions, including Diels-Alder reactions, disulphide, imine bond, boronate ester bond, acylhydrazones, and nanocomposite interactions. An overview of the reaction types, which are discussed herein, is provided in Table 3. Depending on the conditions, dynamic permanent covalent bonds can break and reform. For instance, a disulfide and acylhrazone bonds based self-healing network was developed, where in PEO based hydrogel, disulfide exchange reaction occurs at basic pH, whereas acidic pH is helpful for acylhydrazone exchange reactions (Deng et al., 2012). The material did not show self-healing at neutral pH as the covalent bonds were not dynamic. In general, the broadly investigated temperature triggered reversible diene and dienophile based [4+2] cycloaddition and retro Diels-Alder (DA) mechanism make good candidates for healing networks.
A comprehensive study by Garcia et al. demonstrated the effect of reversible DA bonds in covalently cross-linked networks . The study shows that due to retro-DA reaction at high temperature, an increased mobility in the network leads to the self-healing of the system. However, the required high temperature for DA/retro DA damages the properties of the parent material. A switchable "on" and "off " based photo-switchable DA dynamic network was developed by Hecht et al., where the incorporation of furan crosslinker allows reversible (de-and re-crosslinking) in short and mobile maleimide-substituted poly(lauryl methacrylate) chains ( Figure 5) (Fuhrmann et al., 2016). In general, the common dynamic chemical interactions based self-healing are reversible upon temperature usage, thus limiting their application areas.
The incorporation of nanofillers with this type of selfhealing mechanism is also widely described, as highlighted in Table 4. In particular conductive nanofillers are used to trigger and catalyze the self-repair performance, owing to their superior heat absorption properties and resulting local heating under sunlight and microwave radiation. Indeed, due to their high photo-thermal conversion, conductive nanofillers enable to increase the temperature up to 200 • C in a very short time (Kohlmeyer et al., 2012). Shi et al. developed a dynamic reversible cross-linked organic-inorganic network via DA reaction between poly(styrene-butadiene-styrene) and CNT. The material demonstrated DA based self-healing with complete recovery of the mechanical properties in a time as short as 10 s due to the photo-thermal effect of CNT under laser irradiation (Figure 6)  whereas no self-healing occurred in absence of CNT.

CONCLUSIONS AND OUTLOOK
A large variety of concepts based on physical and chemical interactions have been explored to design soft self-healing materials, each strategy presenting its own advantages and drawbacks. In general, materials based on chemical interactions require an ad-hoc external stimulus to trigger the healing reactions, whereas those based on physical interactions enjoy an autonomous and fast repair. As a matter of fact and as highlighted in many reviews, no standard measurement technique is available to quantify self-healing, so it is difficult to compare the merits of each material in terms of healing efficiency as easily as it is for their strength or toughness.
According to the summarized data, soft self-healing materials based on ionic bonding and hydrogen interactions demonstrate a fast but moderate healing efficiency, whereas a slow but high healing efficiency is observed in case of supramolecular bonding. When considering applications, it is important to keep in mind that mechanical robustness and self-healing properties are often opposite, except in some cases where chemical interactions triggered with external stimulus can be relied upon. However, even if there is no straightforward technique to compare the mechanical properties of soft materials, generally, the modulus followed the usual trend, with covalent bonding > ionic bonding > hydrophobic bonding > supramolecular interactions. The addition of nanofillers enables enhancing the mechanical and other physical properties such as thermal and electrical conductivity (Ding Q. et al., 2018). The addition of conducting nanofiller was found more effective for not only enhancing the conductivity but also mechanical properties of the soft materials. As for selfhealing, conducting nanofillers are used as triggers/catalysts for photo-thermal activation and resulting enhanced selfrepair performance. Laser-induced photo-thermal activation was for instance found to speed up the healing process, but this also led to the partial breakdown of the conductive network, thereby preventing multiple healing. Efficient healing combined with desirable mechanical properties seems at this stage to be reached using a chemical interaction-based concept. Certainly, the selective functionalization and dispersion of nanofillers is quite challenging, and perhaps that is the reason why most of the soft self-healing nanocomposites are nowadays based on physical interactions, rather than on chemical interactions.

AUTHOR CONTRIBUTIONS
RS and BK performed the literature search and tables as well as first drafts. YL, SR, and VM organized the outline and wrote the text. All authors have contributed to the review article.

FUNDING
This work is funded by the Swiss National Foundation Scientific Exchange grant, which financed a visit of SR at EPFL.