Mechanochromic and thermally reprocessable thermosets for autonomic damage reporting and self-healing coatings

Autonomous polymers that report damage prior to loss of function and simultaneously self-heal are highly relevant for preventing catastrophic failures and extending the lifetimes of materials. Here, we demonstrate mechanochromic and thermally reprocessable thermosets that can be used for autonomic damage reporting and self-healing coatings. A mechanochromic molecule, spiropyran (SP), is covalently incorporated into thermoreversible Diels–Alder (DA) cross-linking networks. Mechanical activation of SPs in DA networks is confirmed by computational simulations and mechanical testing. The damaged areas of the polymers change colour, emit fluorescence signals, and completely recover after heat treatment. Because of the thermoreversible covalent networks, these polymers can be recycled up to fifteen times without degrading their mechanical, damage-reporting, or self-healing properties. Our autonomic material systems provide a new way to enhance the lifespans and reliabilities of thermosetting coatings, which also expands the range for practical applications of force-induced chemical reactions in polymers. We first show through computational simulations and experiments that mechanochromic molecules, spiropyrans (SPs), were activated by force in Diels–Alder (DA) reaction-based CANs. Owing to the mechanochromic SP molecules and thermally reversible DA networks, these thermosets indicated damage with colour and fluorescence signals and autonomously repaired it with thermal treatment. While maintaining high solvent resistivity and good mechanical performance, they were reprocessed up to fifteen times without degrading the mechanical, damage-reporting, and self-healing properties. A material that both sends a message that it is damaged and starts to heal itself has been made by researchers in South Korea. A thermoset, which consists of long polymer chains, initially starts in a malleable state. On heating, these chains cross-link, and the material hardens into a permanent arbitrary shape. While this permanence is usually beneficial, it makes repairing damage or recycling difficult. Thermosets have a wide variety of applications, including as coatings and adhesives. Tae Ann Kim from the Korea Institute of Science and Technology, Seoul, and co-workers have created reprocessable and self-healing thermosets. By careful selection of the chemical cross-link, the team were able to create a thermoset that can be reset using heat. When the thermoset is deformed, the affected area changes colour, making the damage instantly visible.


Introduction
Thermosets are chemically cross-linked polymer networks formed by irreversible chemical reactions of soft solids or viscous liquid prepolymers. Owing to their superior mechanical properties, high thermal stabilities, and outstanding chemical resistance, thermosets are used in a wide range of applications, including coatings, adhesives, composites, electronic packaging, etc. 1,2 . As a key element in the modern plastic industry, thermosets comprise 20% of total polymer production today, with worldwide annual production of more than 65 million tons 3 . Although permanent cross-links endow thermosets with desirable properties, their high stability precludes repair, reshaping, and reprocessing of thermosets even at high temperatures; hence, they are among the most difficult materials to recycle 4 .
With increased interest in sustainability and environmental responsibility, dynamic and exchangeable chemical reactions have been investigated for developing reprocessable thermosets known as covalent adaptable networks (CANs) [5][6][7][8] . This type of polymer behaves like a classical thermoset under ambient conditions, but the network topology can be rearranged repeatedly in response to external stimuli. Since every reaction rate is dependent on temperature, thermal treatment is a universal stimulus used to trigger or control dynamic chemical reactions [9][10][11] . In particular, thermoreversible Diels-Alder (DA) reactions have been employed for syntheses of CANs due to their relatively fast kinetics and catalyst-free mild reaction conditions 12,13 . The DA reaction involves a [4 + 2] cycloaddition between an electron-rich diene and an electronpoor dienophile to form a stable cyclohexene derivative.
The use of furan as the diene and maleimide as the dienophile constitutes an excellent combination for CANs since this DA reaction has relatively low coupling and high decoupling temperatures 14 . Specifically, multifunctional maleimides and furans have been used as cross-linking agents with several furan-and maleimide-containing polymers, respectively 15,16 . Therefore, we chose to use DA chemistry as a thermoreversible cross-linking tool for self-healing and reprocessable thermosets.
Autonomous polymers that report damage prior to loss of function and simultaneously self-heal are highly relevant for preventing catastrophic failures and extending the lifetimes of materials, thus ultimately keeping us on the path to sustainability 17,18 . In particular, damage visualization has been achieved by incorporating mechanochromic molecules that change optical properties in response to mechanical stresses into polymeric materials. Spiropyran (SP) is a wellstudied mechanochromic molecule that undergoes a force-induced reversible 6π electrocyclic ring-opening reaction to produce a merocyanine (MC) form 19 . Different strategies have been used to synthesize SP-linked mechanochromic polymers: Homo and block copolymers, including polyacrylates 20,21 and polystyrenes 22,23 , were prepared by using SP as a bifunctional initiator via atom transfer radical polymerization (ATRP). Dihydroxy-terminated SPs were incorporated into polyurethanes 24,25 , polyesters 26,27 and polycarbonates 28 through step-growth or ring-opening polymerization. SP-linked silicone rubbers were fabricated using bisalkene functionalized SP as a cross-linker 29,30 . However, there has been no attempt to integrate SP molecules into CANs, especially those prepared with DA chemistry. Since some DA isomers undergo force-induced retro-DA (rDA) reactions 31 , SP-incorporated DA networks would be useful for studying how two different forcesensitive molecules can be mechanically activated in a single polymeric system.
Here, we demonstrate self-healing and reprocessable thermosets that indicate damage with optical changes involving mechanochromic SPs and thermoreversible DA chemistry (Scheme 1). Linear random copolymers containing furfuryl functionalities are synthesized from bifunctional SP initiators through ATRP. We discuss in detail how polymerization kinetics and comonomer compositions are influenced by reaction conditions. Then, tris-maleimide cross-linkers are mixed with SP-linked linear polymers to prepare a thermally reversible crosslinked network. When the thermoset is mechanically deformed or damaged, the affected areas change colour and emit fluorescence during the SP to MC transition. Upon thermal treatment, the coloured MC reverts to the colourless SP, and reversible DA-rDA reactions reorganize the polymer network, which results in self-healing and reprocessing. Mechanical activation of SPs in DA polymeric networks is investigated with both ab initio steered molecular dynamics simulations and mechanical testing. We also evaluate the self-healing abilities and thermomechanical properties of mechanochromic thermosets after multiple recycling processes.

Materials
Unless otherwise states, all reagents were purchased from commercial source and used as received. Lauryl Nine types of SP-linked copolymers (P) containing SP mechanophores near the chain midpoint were synthesized by changing the molecular weights and FMA molar contents. Ctrl-linked copolymer (C M3 ) was also prepared, and it had a chemical composition and molecular weight similar to those of M3. The synthetic conditions, molecular weight, molar content of FMA incorporated, and glass transition temperature (T g ) are summarized for each sample in Table 1. The ranges of molecular weights for L, M and H were 13 k-19 k, 34 k-38 k and 67 k-77 k, respectively. The ranges of FMA molar contents for 1, 2 and 3 were 2.6-3.0 mol%, 7.1-9.1 mol% and 19-29 mol%, respectively.
The polymerization procedure used for H3 proceeded as follows. LMA (2.8 ml, 9.6 mmol, 320 equiv.), FMA (0.29 ml, 1.9 mmol, 64 equiv.), SP or Ctrl (20 mg, 0.03 mmol, 1 equiv.), CuBr (4.3 mg, 0.03 mmol, 1 equiv.), and copper wire (3 cm) were accurately weighed and transferred to a 50 ml Schlenk flask. After several vacuum and Ar purging cycles, toluene (9.3 ml) and PMDETA (6.3 μl, 0.03 mmol, 1 equiv.) were added to the flask. To remove oxygen, we used the freeze−thaw method three times. The polymerization was started by heating the solution to 80°C under an argon atmosphere. Aliquots were periodically taken with a syringe to determine the reaction conversion and monitor the change in molecular weight. Upon completion of the reaction, the flask was opened to air, and the viscous solution was diluted with THF. The resulting solution was passed through a column of aluminium oxide to remove the copper catalyst, and then the produced was precipitated by addition to methanol. Similar polymerization procedures were adopted for synthesizing different SP-or Ctrl-linked copolymers. Detailed information on each sample is included in the Supplementary material.
Cross-linked polymer using the Diels-Alder reaction (xP and xC M3 ) By mixing P or C M3 with different amounts of trismaleimide cross-linker, cross-linked polymers (xP or xC M3 ) were prepared. The method used for preparation of xL1 with a furan-to-maleimide molar ratio of 1 proceeded as follows. L1 (0.5 g, 0.037 mmol, moles of furan in the polymer = 0.056 mmol) and tris(2-maleimidoethyl)amine (7.3 mg, 0.019 mmol, moles of maleimide in the cross-linker = 0.056 mmol) were dissolved in chloroform (10 wt%) in a 5 ml vial. The mixture was stirred for 1 h and poured onto a Teflon plate. The solution was dried in a vacuum oven and further cured at 50°C for 12 h, resulting in cross-linked polymers. The dried materials were moulded between parallel stainlesssteel plates via compression moulding. A pressure of 4 tons and a temperature of 140°C were applied to the plates for 30 min. The final sheet was cut into barshaped specimens (15 × 5 × 0.7 mm). Similar procedures were adopted to prepare other cross-linked polymers (from xL2 to xH3 and xC).

Simulation methods
We carried out ab initio steered molecular dynamics (AISMD) simulations for five different molecules: three single mechanophore molecules (SP, endo-DA adduct, and exo-DA adduct) and two compound molecules (SPendo-DA compound and SP-exo-DA compound) (Fig. S1a). A force ranging from 0.7 to 3.5 nN was exerted on the molecule for 10 ps at 300 K. We turned on the external force at the beginning of the simulations and maintained the magnitude of the external force during the simulations. We conducted independent AISMD simulations with different applied external forces to investigate the effects of the external forces. In the simulations, constant forces were exerted on two atoms (A1 and A2 in Fig. S1b) to which the polymer chains would be connected. Each force was set to be directed towards a fixed point (P1 and P2 in Fig. S1b). The fixed points were located on the line that passed through the two atoms (A1 and A2). The distance between the point and the atom (A1-P1 and A2-P2) was five times longer than the distance between the atoms (A1-A2). The simulations were performed with the canonical ensemble (NVT) using a Langevin thermostat with a friction constant of 5 × 10 12 s −1 . We employed TeraChem software with the B3LYP functional and 6-31G* basis set 32 . Ten simulations with different initial configurations and velocities were conducted for each kind of molecule. Note that the magnitude of the external force was fixed during each simulation trajectory. We found from our simulations that only two types of bond cleavage reaction occurred within our simulation times: scission of a CO bond of SP and two CC bonds of the DA adduct (each bond is highlighted in Fig. S1a). We calculated the bond lengths of the CO and CC bonds (l CO and l CC , respectively) to examine the bond cleavage reactions. For CC bonds, the lengths of two CC bonds were averaged. A bond cleavage reaction was defined when the distance between two atoms was stretched to more than 2 Å. For the CC bonds of DA adducts, we decided that bond cleavage occurred when both CC bonds broke. For both CO and CC bonds, their bond lengths fluctuated within a range 1 to 1.5 Å before they were broken; therefore, 2 Å seems to be arbitrary but should be sufficiently large to ensure that bond cleavage had occurred. Since CO bond breaking of SP constitutes a ring-opening reaction, the l CO after bond cleavage was 4 Å (inset of Fig. 1a). For the CC bond, however, the DA adduct split into two molecules after bond cleavage. Therefore, l CC kept increasing and varied from 4 to 12 Å (inset of Fig. 1b) during the simulations.
Characterization 1 H NMR and 13 C NMR spectra were recorded on a Bruker Avance III 400 MHz NMR spectrometer with CDCl 3 as the solvent. Spectra were referenced to the residual chloroform solvent peaks ( 1 H NMR: 7.24 ppm, 13 C NMR: 77.23 ppm). Gel permeation chromatography (GPC) measurements were performed on the polymer samples with an Agilent 1260 Infinity II equipped with a 1260 Infinity II refractive index detector (RID) and two PLgel 10 μm MIXED-B columns with a prefilter. Tetrahydrofuran (inhibitor free, HPLC grade, Tedia) was used as the eluent with a flow rate of 1 ml/min. The molecular weights were calibrated by using monodisperse polystyrene standards. Fourier transform infrared (FT-IR) spectra were recorded using a Thermo Fisher Scientific Nicolet 380 spectrometer. The spectra were recorded over the range 4000-650 cm −1 . Optical and fluorescence images were acquired with a Leica DMI6000B camera. Differential scanning calorimetry (DSC) analyses were performed with a Perkin Elmer DSC 4000. The samples were measured over two heating runs from −80°C to 200°C with a heating rate of 10°C/min under a nitrogen atmosphere. Dynamic mechanical analyses (DMA) were performed on a Q800 system from TA Instruments in tension mode and with an amplitude of 0.1% at a frequency of 1 Hz. Storage and loss moduli were measured as a function of temperature from −50°C to 120°C at a heating rate of 5°C/min. Uniaxial tensile tests were also performed on the Q800 system from TA Instruments. Bar-shaped specimens (15 × 5 × 0.7 mm) were subjected to a preload of 0.01 N at a strain rate of 0.05 s −1 . At least three samples were tested. Stress relaxation tests were performed on an MCR 302 instrument from Anton Paar in the 25 mm parallel plate geometry. A normal force of 0.5 N was used in compression mode, and a strain of 1% was applied to the material. Relaxation moduli, G(t), were measured over time at constant temperature (100, 110, 120, 130 and 140°C).

Results and discussion
Ab initio steered molecular dynamics simulations Ab initio steered molecular dynamics (AISMD) simulations were initially used to investigate the mechanochemical reactivities of multiple mechanoresponsive residues, SP and DA adducts, on an atomistic scale (Fig. 1, Movie S1). AISMD is a useful approach for analysing the nonequilibrium reaction dynamics of mechanically induced chemical reactions [33][34][35][36][37] . To reduce computational costs, we prepared artificial molecules with two mechanoresponsive residues connected directly to each other and compared the reaction dynamics of the compound molecules with those of the single mechanophore residues. In addition, the effects of DA adduct stereochemistries, endo and exo isomers, on reaction dynamics were a b R 2 = 0.91 examined for both the compounds and single residues. From 10 simulations with different initial configurations and velocities for each molecule, we calculated the probability of the bond cleavage reaction by counting the number of trajectories in which a cleavage reaction occurred. Φ CO and Φ CC denote the fractions of cleavage reactions for the CO bond in SPs and for both CC bonds in DA adducts, respectively. The lengths (l CO and l CC ) of those bonds were also averaged over the last 400 fs of the total simulation time of 10 ps.
Consistent with previous results for constrained geometries simulated by external force (CoGEF) calculations by Klein et al. 38 , cleavage of the CO bond in a single SP residue occurred with a relatively small external force (1.1 nN) compared to those for reactions of the CC bonds in the DA adduct (3.1 nN for the endo-DA adduct and 3.2 nN for the exo-DA adduct). In compound molecules with multiple mechanophores, however, no CO bond was broken with the same force strength, and all ten trajectories for a single SP exhibited cleavage of the CO bond. A force of at least 1.5 nN was required to completely break the CO bonds (Φ CO = 1.0) in the SP-exo-DA and SP-endo-DA compounds. Cleavage of the CC bonds in DA adducts occurred only when the external force exceeded 2.5 nN. Therefore, we anticipate that SP is mechanically activated prior to DA adduct cleavage within a certain range of external forces, resulting in changes in optical properties due to the SPto-MC transition.
Upon application of a high external force sufficient for mechanical activation of both SP and DA, interesting behaviour was observed for a specific stereoisomeric compound, SP-exo-DA. For example, the CO bond of the SP-exo-DA compound was not cleaved in 20% of the trajectories (Φ CO = 0.8) at F ext = 2.8 nN. For the trajectories in which the CO bond did not break even with F ext = 2.8 nN, we found that breakage of the CC bonds in DA residues had already occurred; the CO bond did not break within our simulation times. On the other hand, in the trajectories where the CO bond broke at F ext = 2.8 nN, cleavage of the CC bond and the CO bond occurred almost simultaneously. In the 20% of trajectories with no CO bond cleavage, the external force might not have been transmitted efficiently to the CO bond of the SP residue due to early cleavage of the CC bonds. This behaviour was not observed for the SP-endo-DA compound. We suggest that the sequence of reactions for mechanophores is dependent on the molecular structure, including the stereochemical structures of complex polymer networks with multiple mechanophores.

SP-linked linear copolymers
Encouraged by the simulation results, we first used ATRP to synthesize SP-linked copolymers comprising lauryl methacrylates (LMA, M 1 ) and furfuryl methacrylates (FMA, M 2 ). The monomer concentration affected the polymerization kinetics and dispersity (Ð) (Fig. 2a). Upon increasing the monomer concentration, polymerization was accelerated (Fig. S2a), but the kinetics deviated from first-order behaviour, and Ð was increased (Table S1) ratio ([M 1 ] 0 :[M 2 ] 0 ) did not influence the polymerization behaviour: the living nature and the rate of polymerization were maintained (Fig. 2b, Fig. S2b). The monomer composition of the synthesized copolymer was calculated with 1 H NMR spectroscopy by using the integrated areas of the signals at 4.9 ppm (-OCH 2protons of FMA) and 3.9 ppm (-OCH 2 -protons of LMA). Since FMA tends to have a reactivity ratio higher than those of alkyl methacrylates 39 S3-S11). Although the FMA content was increased up to 23 mol%, the glass transition temperature (T g ) of the polymer was not changed significantly, as summarized in Table 1 and Fig. S12. By increasing the molecular weights of the polymers, higher T g values were obtained. However, all the linear copolymers had T g values lower than −46°C.

Thermoreversible SP-linked CANs
SP-linked CANs were prepared by dissolving P and cross-linkers in organic solvents, followed by solution casting on a heated glass plate. Formation of the DA adducts was confirmed by their FT-IR spectra (Fig. 3a). After mixing with tris-maleimide cross-linkers, characteristic peaks were observed, including those for the C-H bonds of C=C (700 cm −1 ) and C=O carbonyl groups (1700 cm −1 ) in the maleimide rings (L3 + cross-linker before and after DA) 41 . When the DA reaction was completed (xL3), new absorption peaks appeared for C-O-C ether stretching vibrations (1070 cm −1 ) and C=C double bond stretching vibrations (1774 cm −1 ) of the DA adducts.
DA adducts were thermally cleaved through the rDA reaction, the reverse reaction that formed the initial linear copolymers at temperatures above 100°C 42,43 . When we measured the DSC thermogram of xL3, two endothermic peaks were observed during the first heating cycle (Fig. 3b). These peaks were centred at~120°C and 150°C, indicating rDA reactions of the thermally less stable endo and more stable exo adducts. The heat of the endothermic reaction (ΔH) increased with the amount of cross-linker used, and the value was constant at 24.0 J/g until the molar ratio of cross-linker to P reached 1.0. Therefore, we prepared fully crosslinked samples by adding one equivalent of maleimide derivative per furan moiety.
Solubility tests were conducted to confirm the thermoreversible cross-linking and retro cross-linking reactions (Fig. 3c). The cross-linked polymers were insoluble in toluene at room temperature but dissolved at 110°C since both linear polymers and cross-linkers are soluble in toluene. This is because the rDA reaction occurred above 100°C according to our DSC result. When the temperature was maintained at 50°C for 12 h, the DA reaction proceeded, and cross-linked polymers precipitated from the solution. This solubility change with temperature was reversible.

Autonomic damage-reporting and self-healing coatings
Initially, we used continuous irradiation with ultrasound to look for mechanical activation of SP in linear copolymers in solution. A difunctional control SP (Ctrl) was prepared that could not transfer mechanical forces from polymer chains to the spiro C-O bonds (Fig. S13a). Linear copolymers with similar compositions and molecular weights were prepared from both the SP and Ctrl functional groups. Only the SP-linked polymers (M3) exhibited a visible pink colour after ultrasonication, indicating mechanochemical ring opening of the SP. Ctrl-linked polymers (C M3 ) did not change their colours after application of force, but photochromic behaviour was maintained (Fig. S13b, c).
Cross-linked polymers (xM3 and xC M3 ) were prepared from M3 and C M3 by mixing them with trismaleimide cross-linkers and coating them on glass plates to demonstrate force-induced colour changes in solids. When we scratched each sample with a blunt object, only the xM3 samples showed a purple colour at the damaged area (Fig. 4a). Irradiation with 365 nm UV light changed the colours of both xM3 and xC M3 , indicating that SP and Ctrl were photochromic. Mechanical activation of SP was also confirmed with fluorescence microscopy (Fig. 4b). Strong red fluorescence signals were detected at the scratched areas, suggesting that the SP-to-MC transition was activated by force. The self-healing test was conducted above 150°C, at which temperature the rDA reaction proceeded for the exo adduct. After thermal treatment at 160°C for 1 min, the damaged regions self-healed, and the MC reverted to SP: the purple colour and red fluorescence disappeared. The damage-reporting and self-healing behaviours were repeated multiple times with the same specimen, which makes them especially suitable for use with coatings and related applications. For these kinds of materials, scratchhealing tests provided useful information on the healing efficiency, and the area of the damaged surface area was measured before and after the healing process 44,45 . Using nine types of cross-linked samples (from xL1 to xH3), we evaluated the healing efficiencies and measured the times required for recovery from damage. In most cases, the sample was self-healed by over 90% within a minute, and this was not dependent on either molecular weight or FMA composition. This is because all of the linear polymers had T g values well below the temperature required for the rDA reaction.
Our mechanochromic and self-healing polymers were coated on diverse substrates, including glass, steel, and wood, by using solvent casting (Fig. 4c). To avoid any wetting issues, we chose proper solvents depending on the substrates. Once the materials were fully cured, the coated areas were scratched using tweezers. The scratched regions turned purple and completely healed within a minute after thermal treatment (Movie S2). We anticipate that our polymers can be applied in versatile coating materials that require self-reporting and self-healing capabilities. to determine the effects of molecular weight and FMA molar content. E′ refers to the storage modulus, and tan δ refers to the loss tangent. c Normalized relaxation modulus (G(t)/G 0 ) for xH3 as a function of time at constant temperature. The relaxation times (τ*) were defined when G(t)/G 0 was equal to 1/e (dashed lines). d Arrhenius plots obtained from τ* at different temperatures. The activation energy (E a ) for the rDA reaction was calculated from a linear fit. e, f Representative engineering stress (σ) and strain (ε) curves for xP samples.

Thermomechanical properties and reprocessability
We analysed the thermomechanical properties of xP by using dynamic mechanical analyses (DMA) in tension mode (Fig. 5a, b). With increasing FMA molar content (from 1 to 3), the storage moduli (E′) in the glassy and rubbery regions increased while the height of the loss tangent (tan δ) decreased, implying that the material became more elastic due to the increased number of cross-linking sites (Fig. 5a, Fig. S14). T g values, as determined from the peak of tan δ, for xH1 (0°C), xH2 (23°C) and xH3 (50°C) increased as well. Below the temperature required for the rDA reaction (120°C), a single glass transition was observed that became broader with higher FMA content, implying an increase in the inhomogeneity of the matrix. We could not find any trends for E′ and tan δ when the molecular weight of xP was changed with constant FMA molar content (Fig. 5b, Fig. S15). For instance, x1 and x2 had higher moduli with lower molecular weights, while the reverse trend was observed for x3. The peak heights and shapes of tan δ were similar, but T g decreased with increasing molecular weight.
The stress relaxation rates of xH3 were measured as a function of temperature to determine the activation energy (E a ) of the rDA reaction (Fig. 5c). The relaxation time at each temperature was obtained by using Maxwell's model, and the value of E a was obtained with the Arrhenius equation (see the Supplementary material for equations). The value of E a was calculated as 105 kJ mol −1 , which is somewhat above the range of values previously reported (88-95 kJ mol −1 ) for rDA reactions run within a similar temperature range (Fig. 5d) 46 . This increase might be caused by the difficult diffusion of bulky tris-maleimide cross-linkers in viscous polymeric media.
The engineering stress (σ) and strain (ε) responses of xP were characterized under the application of uniaxial tension. Increasing the FMA molar content with similar molecular weights increased the brittleness of the materials, which is consistent with the DMA results: the Young's modulus (E) increased and the elongation at break (ε max ) decreased with increasing FMA molar content (Fig. 5e, Fig. S16). The tensile strength (σ TS ) values for x1 and x2 were similar, but the highest value was observed for x3. Changes in the molecular weight with a fixed FMA content also affected the tensile properties of xP (Fig. 5f, Fig. S17). We anticipate that there is a range of molecular weights that would show optimal mechanical properties based on the amount of FMA. Unexpectedly, we did not observe colour changes under tension for any of the xP samples, which is inconsistent with the behaviour of other SP-linked polymeric systems.
Our polymers can be reprocessed several times with compression moulding, confirming the robust nature of Diels-Alder chemistry. We performed up to fifteen recycling processes with xH2 samples and measured E, σ TS and ε max with tensile testing (Fig. 6). With a series of t tests, we found that the differences in the stress-strain curves for each sample were not statistically significant. Furthermore, there was no difference in E, σ TS and ε max for the recycled and pristine samples at the 95% confidence level. Specifically, fifteen recycled xH2 samples (E = 343 ± 49 MPa, σ TS = 72.7 ± 6.7 MPa, ε max = 27.6 ± 5.5%) exhibited mechanical properties a b Fig. 6 Mechanical properties of pristine and recycled xH2. a Representative engineering stress (σ) and strain (ε) curves for virgin, 5x recycled, 10x recycled, and 15x recycled xH2 specimens. b Elastic modulus (E), tensile strength (σ TS ), and elongation at break (ε max ) for xH2 as a function of recycling runs. Reported values and error bars represent the average and one standard deviation, respectively. that were experimentally identical to those of the pristine sample (E = 321 ± 42 MPa, σ TS = 63.2 ± 3.8 MPa, ε max = 24.6 ± 3.0%). These specimens retained their mechanochromic and self-healing capabilities (Fig. S18).

Conclusions
We demonstrated self-healing and thermally reprocessable thermosetting polymers that change colour and emit fluorescence in response to mechanical stimuli. With ab initio steered molecular dynamics simulations, we confirmed mechanical activation of spiropyrans (SPs) in thermoreversible and force-sensitive Diels-Alder (DA) networks. Nine cross-linked DA networks containing SP were prepared by changing the molecular weights and the cross-linking densities of linear polymers. These polymers exhibited purple colours and red fluorescence after compression or scratching. The damaged areas were self-healed, and the colour had disappeared within a minute after thermal treatment. In addition, the cross-linked polymers showed reversible solubility changes as well as thermal reprocessability. With up to fifteen reprocessing cycles, the thermosets retained their mechanical, damage-reporting, and selfhealing properties. We envision that our autonomic polymers can be applied as universal coatings to enhance the lifespan and reliability of products by warning of damage with colour and then undergoing self-healing.