Conditional repair by locally switching the thermal healing capability of dynamic covalent polymers with light

Healable materials could play an important role in reducing the environmental footprint of our modern technological society through extending the life cycles of consumer products and constructions. However, as most healing processes are carried out by heat alone, the ability to heal damage generally kills the parent material's thermal and mechanical properties. Here we present a dynamic covalent polymer network whose thermal healing ability can be switched ‘on' and ‘off' on demand by light, thereby providing local control over repair while retaining the advantageous macroscopic properties of static polymer networks. We employ a photoswitchable furan-based crosslinker, which reacts with short and mobile maleimide-substituted poly(lauryl methacrylate) chains forming strong covalent bonds while simultaneously allowing the reversible, spatiotemporally resolved control over thermally induced de- and re-crosslinking. We reason that our system can be adapted to more complex materials and has the potential to impact applications in responsive coatings, photolithography and microfabrication.

F uture technologies require smart materials with advanced properties including responsiveness to external stimuli and in particular the ability to autonomously repair inflicted damage [1][2][3] . Such ability to heal defects is typically achieved by the release of microencapsulated healing agents 4,5 or by the introduction of reversible connections in the polymer architecture, either in a non-covalent fashion, such as in supramolecular polymers 2,6 , or by using dynamic covalent chemistry [7][8][9] . One particularly clever approach involves the implementation of reversible polymer networks 10,11 as they offer outstanding mechanical strength and thermal properties that are readily modified by the nature of the connecting dynamic bonds and the crosslinking density. In this context, the reversible Diels-Alder reaction of furans with maleimides is extremely popular 11,12 as it does not require additional catalyst or produce by-products leading to static covalent crosslinks at ambient conditions. At elevated temperatures these polymer networks, however, become dynamic and reorganization within the material to repair inflicted damage is enabled. In these thermoresponsive, reversibly cross-linked polymers damage repair is solely triggered and driven by heat acquiring the ability to heal at the expense of the desirable thermal and hence mechanical properties of a static polymer network. Consequently, conditional healing of the materials under load and in specific areas is difficult to achieve 13 .
Other materials try to circumvent this issue by activation of their healing or repairing ability through chemical signals, such as pH (ref. 14) or redox changes 15 , or physical stimuli, such as mechanical stress 16 . However, as it would be highly attractive to selectively address and reversibly switch the dynamic covalent bond 'on' or 'off', light arguably presents itself as the ideal stimulus: It offers the possibility to activate molecules with a high spatiotemporal as well as energetic resolution without the need to invasively alter the material properties thus rendering photoswitches the ideal molecules to locally control thermal healing behaviour. In stark contrast, solely photo-driven approaches require continuous light irradiation to power the healing process [17][18][19][20] .
Our conditional healing approach relies on a set of photoswitchable reactants 21 whose ability to undergo a thermally reversible crosslinking reaction (X ON þ P"X@P ON )-that is, to connect and disconnect polymer chains (P)-can be controlled by light to switch between a dynamic (ON) and a static system (OFF) (Fig. 1a). By illumination with ultraviolet-light, both the free crosslinker X ON as well as the cross-linked polymer X@P ON can be converted to their respective locked states (X OFF and X@P OFF ) from which the dynamic covalent reaction is inhibited. From there, visible light can selectively induce the re-activation of the polymer's dynamic character leading to a reversibly switchable dynamic covalent material. We designed our system based on photoswitchable furans since Branda and coworkers 22,23 as well as our own group 24  (a) Schematic representation of a photoswitchable crosslinker X ON in its reactive state and polymer P carrying complementary reactive groups forming the dynamic covalent polymer network X@P ON , which can be decrosslinked on heating. On illumination with ultraviolet-light the crosslinker X as well as the polymer network X@P are reversibly transformed into their inactive state (X OFF and X@P OFF , respectively), where both crosslinking as well as de-crosslinking are inhibited. (b) Chemical structures of the tetrafurylsubstituted diarylethene crosslinker, which in its activated form X ON reacts with maleimide-functionalized poly(lauryl methacrylates) via the Diels-Alder reaction (for simplicity only two of four crosslinking units are shown). On heating the network is de-crosslinked due to the retro Diels-Alder reaction. Illumination with ultraviolet-light induces 6p-electrocyclization to the ring-closed isomers of the deactivated crosslinker X OFF and the polymer X@P OFF , respectively. This process can be reversed and the crosslinker as well as the network can be activated by illumination with visible light.
furan-containing diarylethene photoswitches employed in these works are especially well suited for this purpose as they guarantee the bistability of the switchable material and can be addressed selectively and independently using ultraviolet and visible light 26 .

Results
Design and synthesis of crosslinker and linear polymer. For efficient crosslinking, more than one reactive furyl moiety is necessary and hence we utilize a tetrafunctional motif X inspired by Wudl and coworkers 11 minimizing the amount of crosslinker and hence optical density to warrant maximum light penetration. This is an intrinsic advantage over photocontrollable linearly connected polymer chains where the photochromic compounds form the polymer backbone 23 . Moreover, the connection of the diarylethene to the pentaerythritol core via electron-withdrawing ester groups guarantees high fatigue resistance and photochemical efficiency 27 , while retaining the beneficial spectral separation of the different isomers (Fig. 1b, for synthesis see Supplementary Methods). A further benefit over already reported systems 23 is the ability to switch 'off' the unreacted crosslinker adding the possibility to activate the dynamic bond on demand. Complementary reactive groups are provided via well-defined short linear polymer chains of maleimide functionalized poly(lauryl methacrylates) P (M n ¼ 6,000-8,600 g mol À 1 ) synthesized by atom transfer radical polymerization with narrow dispersity (Ð ¼ 1.17-1.26) and with a variable amount of maleimide that can be adjusted modularly to achieve the best crosslinking density (for synthesis see Supplementary Methods) 28 . The enhanced mobility necessary for a surface rearrangement and effective reflow of the material-in particular in thin film coatings-is accomplished by introduction of dodecyl side chains in the polymer backbone warranting a low glass transition temperature of T g ¼ À 44°C (Supplementary Fig. 1) 29,30 and additionally providing optical transparency above 350 nm ( Supplementary Fig. 2).
Controlling the healing ability in scratched thin films. Local control over the thermal healing process was investigated using scratched thin films of X@P ON (Fig. 2) prepared from X ON carrying four furyl groups and P (11 mol% maleimide content) mixed in a ratio to assure 0.7 equivalents of furyl groups per maleimide unit to minimize the optical density. The crosslinking process was determined to yield roughly 88% conversion of the furyl groups by ultraviolet/visible spectroscopy (Supplementary Generation and characterization of a static polymer network. The selective transition from the dynamic to the static state ( Fig. 2a,b) was carried out by illumination of the scratched crosslinked network X@P ON with ultraviolet-light (365 nm LED) through a photomask generating X@P OFF locally. The locked network X@P OFF clearly contains ring-closed crosslinker as indicated by the colouration of the exposed area. The formation of X@P OFF is confirmed by solid state ultraviolet/visible spectroscopy where the characteristic absorption band of the ringclosed Diels-Alder adduct-as compared with the respective small molecule reference compound in solution ( Supplementary Fig. 11, Supplementary Note 4)-can be retrieved in the illuminated area (Fig. 3a). Moreover, locking X@P ON to X@P OFF can be followed by Fourier transform infrared (FT-IR) spectroscopy, where a shift of the characteristic out of plane C-H bending mode of the vinylic group in X@P ON at 868 to 854 cm À 1 corresponding to the allylic group in X@P OFF is clearly visible (Fig. 3b, blue dashed lines to yellow dashed lines) 31 . Interestingly, on ultraviolet-illumination the mean polymeric mesh size increases from 61±0.10) nm for X@P OFF as determined by SAXS measurements (Fig. 3c; Supplementary Figs 5, 12 and 13). At first, this finding appears to be counterintuitive but we attribute it to a more efficient photochemical ring-closure in the smaller meshes due to the higher optical density of the photoswitch. This leads to preferential 'disappearance' of the smaller meshes as they fall below the SAXS lower size detection limit of 40.7-0.8 nm effectively widening the mesh size (Supplementary Table 1 and Supplementary Note 3).
Conditional repair of the polymer network. In addition to the revealing visual comparison of the unlocked, dynamic and the locked, static networks' response to heat with healing occurring only in the non-illuminated area, (Fig. 2b,c) the structural molecular basis of these macroscopic changes can be monitored by FT-IR spectroscopy. While heating of X@P ON above 120°C results in a significant decrease in the vinyl group's C-H bending signal clearly indicating the de-crosslinking process, (Fig. 3b, blue dashed line to green solid line) heating X@P OFF to the same temperature retains the spectral features demonstrating the successful inhibition of the retro Diels-Alder reaction by light (Fig. 3b, yellow dashed line to yellow solid line). Examining the thermo-rheological behaviour of X@P ON reveals a typical drop of the shear storage modulus G 0 below the loss modulus G 00 above 100°C that is well documented for Diels-Alder cross-linked polymers 32,33 and verifies the dynamic de-crosslinking process (Fig. 3e, blue to green lines). In stark contrast, X@P OFF retains its static viscoelastic properties on heating, that is, G 0 4G 00 regardless of the temperature, providing additional proof for the successfully inhibited retro Diels-Alder reaction on the ultraviolet-  illuminated side (Fig. 3e, yellow lines, Supplementary Fig. 14).
Monitoring of a number of distinct vibrational modes in comparison to small reference molecules by FT-IR and further SAXS as well as differential scanning calorimetry (DSC) measurements strongly underline this result (Supplementary  Figs 7, 12, 13 and 15-17, Supplementary Tables 2 and 3

, and Supplementary Notes 3 and 5).
Restoration of the polymer network's dynamic character.
Unlocking X@P OFF to X@P ON , thus restoring the polymer's dynamic character and its healing ability, (Fig. 2c,d) is realized by illumination with visible light (460 nm LED) and can easily be verified by the naked eye through the apparent decolouration of the material. Due to the rather slow ring-opening, no special precautions have to be taken when handling the locked network X@P OFF under ambient, low intensity sunlight over a time frame of several hours (Supplementary Fig. 18 and Supplementary Note 6). The ring-opening process can be followed in more detail by monitoring the restoration of the vibrational band at 868 cm À 1 using FT-IR spectroscopy (Fig. 3d, yellow dashed line to blue dashed line) as well as by the characteristic changes in the ultraviolet/visible spectrum ( Supplementary Fig. 19). The observed recovery of the polymeric mesh's initial correlation length of X@P ON as monitored by SAXS adds further evidence to this step (Fig. 3c, Supplementary Figs 13  Re-established healing capability. Besides the clearly visible vanishing of the scratch on heating of the re-activated dynamic material (Fig. 2d,e), the regenerated healing ability of X@P ON is once again accompanied by a strong decrease of the indicative signal at 868 cm À 1 in the FT-IR spectrum (Fig. 3d, blue dashed lines to green solid lines) and a full widening of the polymeric mesh from x 2 ¼ (1.27 ± 0.11) nm to x 2 ¼ (1.93 ± 0.12) nm (Fig. 3c, Supplementary Fig. 20, Supplementary Table 5 and Supplementary Note 3). The light-induced local healing ability strongly suggests a chemical de-crosslinking/re-crosslinking process rather than physical melting to be responsible for the observed behaviour. This is further corroborated by rheology experiments at varying temperatures that provide insight into the kinetics of the (retro) Diels-Alder reactions, illustrating the competing effect of thermally driven de-crosslinking and thermally activated re-crosslinking ( Supplementary Figs 8 and 21).
Reversibility of all steps. The material undergoing the lightcontrolled locking and de-locking cycle fully corresponds to X@P ON as generated by direct conversion (Fig. 2a,e) in every analytical method we employed, that is, optical (ultraviolet/ visible, FT-IR), structural (SAXS), mechanical (temperaturedependent shear rheology) and thermal (DSC) properties. This highlights the reversibility of all steps involving illumination with ultraviolet and visible light as well as thermal treatment in the locked and unlocked state.

Discussion
Our photoswitchable dynamic covalent material consists of linear polymethacrylate chains substituted with maleimide units that are connected to a network by a multivalent light-responsive furan crosslinker. The incorporation of the latter allows us to transform specific areas reversibly and robustly from a thermally healable dynamic to a static 'locked' polymer network and vice versa simply by locally applying light of different wavelengths. Extensive rheology, DSC, ultraviolet/visible and SAXS measurements provide sound evidence that the observed conditional healing truly has its origin in the formation and scission of covalent bonds via the (retro) Diels-Alder reaction, as expected on the basis of the pioneering work by Wudl 11 and others 12 .
We believe that the introduction of this additional switching stimulus solves one of the major drawbacks of healable materials that commonly attain their healing ability at the expense of their thermal integrity. As opposed to the method described by Rowan, Weder, and coworkers 17 , this photoswitchable locking and unlocking moreover avoids continuous irradiation and thus light-induced material degradation. Moreover and in contrast to solely photohealable polymers, the combination of light activation with subsequent thermal healing provides better healability, presumably due to an enhanced material mobility necessary for repair 20 . While presented here in the context of thin film applications and functional coatings, we reason that our reactant system can be adapted to numerous classes of more complex, responsive polymer systems. The associated external control over local dynamic crosslinking and hence healing ability should aid the design of programmable, repairable soft materials for a multitude of future applications. This could include macroscopic properties, such as in healable paints, as well as microscopic features, such as in latent resists that can be carried through various processing steps until being activated to offer control over nanofabrication.

Methods
General remarks. For syntheses and not previously mentioned characterization of crosslinker X ON , the maleimide-functionalized monomer MIMA and furyl-protected, inactive linear polymer fpP as well as non-protected, active linear polymer P see Supplementary Figs 22

and 23 and Supplementary
Methods.
General procedure for preparation of X@P ON and X@P OFF . Copolymer P (1.0 equiv. maleimide side chains) and crosslinker X ON (0.7 equiv. furan termini) were dissolved in a minimum amount of dry THF in a 2.5 ml vial and subsequently drop-casted onto a glass slide positioned in a Schlenk flask under an argon atmosphere followed by subsequent evacuation. The evacuated Schlenk flask was heated at 130°C for 90 min to evaporate remaining solvent and to anneal the polymer mixture, followed by thermal crosslinking at r.t. for additional 16 h in the evacuated Schlenk flask. For rheology, drop-casting was performed directly onto steel plates. Note that polymer network preparation has been optimized for thin films and no free-standing films (which would allow a classical tensile test) have been obtained. After thermal crosslinking, half of the bulk material was irradiated with a Roithner 365 nm-LED XSL-365-5E at 20 mA and 4.2 V LED in a distance of 3 cm orthogonal to the sample and used as the locked polymer network sample X@P OFF for further analysis. Furthermore, the locked and heated polymer film was irradiated with a LED Engin 460 nm Blue LED Emitter LZ4-00B208 at 2-3 mA and 12 V for 105 min to further investigate the reestablishment of the healability.
General procedure for scratching tests of X@P ON and X@P OFF . Scratches were done with a scalpel in a controlled manner in the mm-scale, followed by masking half of the scratch and cross-linked polymer film X@P ON (prepared as described above) with aluminium foil. Illumination of the unmasked area was performed in a Schlenk flask under an argon atmosphere for 30 min by adjusting the abovementioned Roithner 365 nm in a distance of 3 cm orthogonal to the glass slide to form X@P OFF . After removal of the aluminium foil, the glass slide was heated on a Deben Enhanced Coolstage from ambient temperatures to 124°C in a specimen chamber vacuum (B30-50 Pa) of a SEM TM-1000 control unit remaining at that temperature for 5 min. For reestablishment of the healability, the polymer film was irradiated with the above-mentioned LED Engin 460 nm-Blue LED Emitter for 105 min, followed by another heating process (same conditions as for locking the polymer). Optical micrographs as well as FT-IR spectra were recorded after each step.
Instrumentation. Optical micrographs were either acquired on a Bruker A670 Hyperion FT-IR microscope in the reflection mode with a visible objective, 40-times magnified ( Â 10 ocular, Â 4 vis-objective, for micrographs in main article) or with the optical detection unit of the atomic force microscopy (AFM) facility (for micrographs presented in the SI  Supplementary Figs 8 and 21b. Note that compression or tensile measurements could not be performed due to the film's high viscoelasticity and creep even at low temperature. SAXS measurements of X@P ON and X@P OFF were performed in a solid sample holder with a Kratky-type instrument (SAXSess from Anton Paar, Austria) at temperatures of (21±1)°C and (120 ± 2)°C. The SAXSess has a low sample-to-detector distance (0.309 m), which is appropriate for short measurement times of 10 min. The measured intensity was corrected by subtracting the intensity of the empty sample holder with a 30 mm-thick aluminium foil. The scattering vector is defined in terms of the scattering angle y and the wavelength of the radiation (l ¼ 0.154 nm): thus q ¼ 4pn/l siny. Deconvolution (slit length desmearing) of the SAXS curves was performed with the SAXS-Quant software. Curve fitting was conducted with the software SASfit. For solid state ultraviolet/visible spectroscopy of X@P ON and X@P OFF 3-5 ml of a solution of free crosslinker X ON (0.7 equivalent furan per maleimide unit) and the respective polymer in degassed THF were spin-coated on 1 Â 1 cm quartz glass plates (thickness of plates: 1 mm; polymer film thickness approx. 2.0 mm, determined by AFM measurements in tapping mode). Spin coating was performed at a rotation speed of 100-150 rps with a spin coating time set to 60 s with a KLM spin coater SCC-200 from SCHAEFER Technologies Corporation (Langen, Germany) at room temperature. Thermal crosslinking was carried out for 16 h at r.t. Irradiation was performed directly in a Varian Cary 50 ultraviolet/visible spectrophotometer equipped with a Peltier thermostated cell holder at 25±0.05°C by adjusting the LED in a distance of 1 cm orthogonal to the quartz glass plate in the sample holder. A Roithner 365 nm-LED XSL-365-5E for ring-closing at 20 mA and 4.2 V and a LED Engin 460 nm-Blue LED Emitter LZ4-00B208 for ring-opening at 2-3 mA and 12 V were employed, both driven by a GW Instek GPD-3303S linear DC power supply.
Data availability. All data that support the findings of this study are available from the corresponding author on reasonable request.