A Thermally Re-mendable Cross-Linked Polymeric Material

See allHide authors and affiliations

Science  01 Mar 2002:
Vol. 295, Issue 5560, pp. 1698-1702
DOI: 10.1126/science.1065879


We have developed a transparent organic polymeric material that can repeatedly mend or “re-mend” itself under mild conditions. The material is a tough solid at room temperature and below with mechanical properties equaling those of commercial epoxy resins. At temperatures above 120°C, approximately 30% (as determined by solid-state nuclear magnetic resonance spectroscopy) of “intermonomer” linkages disconnect but then reconnect upon cooling, This process is fully reversible and can be used to restore a fractured part of the polymer multiple times, and it does not require additional ingredients such as a catalyst, additional monomer, or special surface treatment of the fractured interface.

In past decades, highly cross-linked polymers have been studied widely as matrices for composites, foamed structures, structural adhesives, insulators for electronic packaging, etc. (1,2). The densely cross-linked structures are the basis of superior mechanical properties such as high modulus, high fracture strength, and solvent resistance. However, these materials are irreversibly damaged by high stresses (3,4) due to the formation and propagation of cracks. The latter lead to dangerous loss in the load-carrying capacity of polymeric structural engineering materials (5,6).

The exploration of re-mending and self-healing of polymeric materials has become increasingly more exciting in the recent past, culminating in a genuinely self-healing material (4). The hot plate welding and crack healing of thermoplastics, where intermolecular noncovalent interactions (chain entanglements) at the interface are responsible for mending, have been well established (7–9). Small molecules–induced crack healing has also been studied for thermoplastics (10,11), and a composite of a linear polymer with a thermoset that has crack-healing ability was reported (12). The concept of self-repair was introduced to heal cracks by embedding hollow fibers that can release repair chemicals when a crack propagates (7, 13, 14). The most recent report of crack self-healing (autonomic healing) of an epoxy resin consists of a clever use of catalytic network formation of an encapsulated add-monomer, which is held within a capsule embedded in the epoxy matrix (4). But questions remain concerning the long-term stability of the catalyst and the ability of the material to self-heal multiple times. We describe a truly “re-mending” material consisting of a highly cross-linked, transparent polymer that exhibits multiple cycles of autonomic crack mending with simple, uncatalyzed thermal treatment and forms covalent bonds at the interface of the mended parts. In contrast, in hot plate welding and crack healing of thermoplastics, only intermolecular, noncovalent interactions (chain entanglements, hydrogen bonding, etc.) are responsible for the mend and no new covalent bonds are formed between the mended parts.

Thermally reversible reactions, particularly the Diels-Alder (DA) reaction, for cross-linking linear polymers have been pioneered and studied by Kennedy and Wagener during the past two decades (15–21). To the best of our knowledge, as shown in all the reports of thermoreversible polymers, cross-linkable and reversible groups were attached to linear polymer backbones, but the connections of the cross-linkers to the polymer backbones were not reversible (15–21). In contrast, we report a macromolecular network formed in its entirety by reversible cross-linking covalent bonds.

A thermally reversible DA cycloaddition of a multi-diene (multi-furan, F) and multi-dienophile (multi-maleimide, M) was used to prepare a polymeric material (Scheme 1). Monomer 1(4F) contains four furan moieties on each molecule, and monomer2 (3M) includes three maleimide moieties on each molecule (22). A highly cross-linked network (polymer3, 3M4F) can be formed via the DA reaction of furan and maleimide moieties, and thermal reversibility can be accomplished by the retro-DA reaction (23).

A dichloromethane solution of monomers 1 and2 (with stoichiometric F:M ratio) was poured into a glass mold or cast on a quartz plate. Vacuum evaporation of the solvent at room temperature followed by heating produced the macromolecular solid. Polymerization at different temperatures was studied by ultraviolet (UV) spectroscopy. Monoliths of polymeric material were obtained by breaking the glass mold, whereas thin films were produced when cast on quartz substrates. Because the polymerization process involves the formation of DA adducts, the efficiency of the DA reaction can be used to study the degree of polymerization. According to the literature (19), the extent of the DA reaction for a similar system could be followed quantitatively by UV spectroscopy, and the degree of cross-linking as a function of reaction temperature is shown in Fig. 1. At 24°C, polymerization of up to 60 to 70% required 5 days. The polymerization–cross-linking process is much faster at higher temperatures, reaching “completion” (95% ± 5%) in just 3 hours at 75°C.

Figure 1

Polymerization degree versus time. (□) 24°C; (▾) 45°C; (▵) 60°C; (▪) 75°C.

Differential scanning calorimetry (DSC) showed that the polymerization process is exothermic by 23 kcal/mol (per DA adduct) (24). Assuming the principle of microscopic reversibility applies, this result implies that upon heating the retro-DA reaction is an accessible reaction pathway that is preferred over (nonreversible) bond-breaking degradation reactions (25) in the polymer network.

The bulk polymeric material is hard and fully transparent, with a density of 1.37 g/cm3 and an index of refraction of 1.56 at 24°C. On the basis of a nondestructive ultrasonics technique (26), the Young's modulus and Poisson ratio were determined to be 4.72 GPa and 0.349, respectively. From the results shown in Table 1, it is clear that the mechanical properties of this polymer are in the range of commercial, state-of-the-art, cross-linked epoxy resins and unsaturated polyesters (1, 32).

Table 1

Mechanical properties of polymer 3 compared with epoxy resins [data taken from p. 250 of (1)] and unsaturated polyesters (Unsat. poly.) [data taken from p. 250 of (1)]. Methods follow the guidelines of the American Society for Testing and Materials (ASTM). Dashes indicate data not reliable by method used.

View this table:

Solid state reversibility (27), was tested via a series of heating cycles that consisted of heating fully polymerized films to different temperatures on quartz substrates and then quenching in liquid nitrogen. After the heat-quench procedure, the samples were heated at 80°C, causing repolymerization and/or cross-linking of the films. The results of five such cycles are shown in Fig. 2. About 25% of the DA adducts disconnected after thermal treatment at 150°C, but only ∼12% did so after treatment at 130°C (first half cycle, left side of Fig. 2). Under these conditions, the sample was observed to be irreversible below 120°C. Prolonged storage of the films at 80°C afforded full reconnection of the DA partners. Clearly, the simplest interpretation of these five continuous cycles demands that the DA and retro-DA reactions are reversible in this extended solid.

Figure 2

Thermal reversibility of polymer 3. (A) to (B), 130°C for 25 min, followed by quenching in liquid nitrogen. (B) to (C), (D) to (E), (F) to (G), and (H) to (I), 80°C for 1 hour. (C) to (D), (E) to (F), (G) to (H), and (I) to (J), 150°C for 15 min, followed by quenching in liquid nitrogen.

As shown in Fig. 3, samples with different thermal history were studied by temperature-modulated DSC (TMDSC). This type of thermal analysis is used to determine and distinguish between (i) fully reversible heat effects that are related to the heat capacity change and (ii) time-dependent irreversible heat effects that are related to kinetics (28, 29). InFig. 3A, a fully polymerized sample, which was cooled from 145° to 25°C in 24 hours, showed only an endothermic peak at 124°C. The shape is typical for physical aging effects (30). A sample that was kept at 145°C for 25 min and quenched in liquid nitrogen showed an exothermic peak from 60° to 120°C on the nonreversing heat flow curve (Fig. 3B). These results indicate that the quenched sample contains disconnected furan and maleimide moieties, and, as the molecular chains gain enough mobility at increased temperature, they reconnect efficiently. At temperatures above 120°C, endothermic peaks in both nonreversing heat flows (due to the retro-DA reaction in the network) were observed.

Figure 3

Modulated DSC data of polymer 3(scanning rate, 5°C/min). (A) Sample was cooled from 145° to 24°C in 24 hours. (B) Sample was heated to 145°C for 25 min and then quenched in liquid nitrogen. “1” marks total heat flow; “2,” reversing heat flow; “3,” nonreversing heat flow.

Solid state 13C-NMR spectroscopy was also used to study the thermally reversible DA reaction of the macromolecular solid. Solution 13C-NMR spectroscopy of the 4F monomer revealed resonances for positions 2, 3, 4, and 5 of the furan moieties at 149.5, 110.6, 110.6, and 143.2 parts per million (ppm), respectively. Upon cycloaddition, the C-3 and C-4 positions shift to around 138 ppm, and the C-2 and C-5 resonances shift up-field (82.4 and 90.7 ppm). To confirm our assignments, the DA adduct of furfuryl acetate with maleic anhydride was used as a model system (see structures with Fig. 4). Solid state 13C-NMR spectra of polymer 3 as a function of thermal cycling are shown in Fig. 4. Figure 4A is a spectrum of the fully polymerized extended solid with a clear absence of peaks at 110 and 150 ppm. Heating the polymer to 145°C for 25 min and then quenching in liquid nitrogen produced material with the solid state 13C-NMR spectrum shown in Fig. 4B. The ∼ 110 and ∼150 ppm resonances are visible, indicating the presence of unreacted furan moieties. The same sample was heated to 120°C and cooled down to 70°C in 24 hours to repolymerize. Its solid state 13C-NMR spectrum is shown in Fig. 4C. Because the latter is essentially identical to the spectrum in Fig. 4A and no peaks due to furan moieties were observed, the simplest interpretation demands that, within the experimental error of solid state NMR spectroscopy, all the disconnected linkages had reconnected.

Figure 4

Solid State 13C-NMR spectra of (A) a fully polymerized sample, (B) a sample heated to 145°C and then quenched in liquid nitrogen, and (C) a repolymerized sample (cooled from 120° to 70°C in 24 hours).

Because the bond strength between diene and dienophile of the DA adduct is much lower than all the other covalent bonds, the retro-DA reaction should be the major reason for crack propagation. In principle, when the sample is reheated, the furan and maleimide moieties should reconnect and the cracks, or fractures, should mend.

To determine the fracture-mending efficiency of this polymer, tests were performed with the use of compact tension test specimens (4, 12, 31) (Fig. 5). A sharp pre-crack was created in the tapered samples by gently tapping a fresh razor blade into a machined starter notch. Application of a load in the direction perpendicular to the pre-crack, led to a fracture of the specimen because it was not possible to stop the crack propagation. After structural failure, the two pieces were matched as closely as possible and were held together with a clamp, treated at 120° to 150°C under nitrogen for about 2 hours, and cooled to room temperature. Photographs of the mending effect of a typical specimen are shown in Fig. 5, B and C. Before mending, the interface of the crack was highly reflectant (Fig. 5B). After thermal treatment, only a slight interface mismatch was observable (Fig. 5C), indicating mending (healing) of the interface.

Figure 5

Mending efficiency of polymer 3. (A) Mending efficiency obtained by fracture toughness testing of compact tension test specimens. Values for the original and healed fracture toughness were determined by the propagation of the starter crack along the middle plane of the specimen at the critical load. (B) Image of a broken specimen before thermal treatment. (C) Image of the specimen after thermal treatment. (D) SEM image of the surface of a healed sample: the left side is the as-healed surface and the right side is the scraped surface. (E) Enlarged image of the boxed area in (D).

Scheme 1

Scanning electron microscope (SEM) images of a healed sample are also shown in Fig. 5, D and E. Before the images were taken, Cr (3 nm) and Au (30 nm) were deposited onto the surface via electron-beam evaporation. Shown on the left side of Fig. 5D is the scar of the mended interface; the right side is where the surface was scraped with a fresh razor blade. Because the two pieces that were mended could not be matched perfectly when held together, the scar on the healed surface is clearly visible as a white line. However, the surface scraped with the razor blade reveals that the fracture healed almost completely to produce a homogeneous material with a few minor defects, suggesting remarkable mending efficiency. Figure 5E is an enlarged image of Fig. 5D. The latter shows the interface between the scarred section and the scraped section, and it gives a clearer, more dramatic view of the mending effect.

Fracture tests were also carried out in an effort to quantify the healing efficiency (4). Representative load-displacement curves for a polymer specimen are plotted in Fig. 5A, showing recovery of about 57% of the original fracture load. Not surprisingly, the accurate replacing of the two halves and thermal treatment temperatures affected the mending efficiencies. At 150°C, an average mending efficiency of about 50% was achieved, whereas at 120°C the average value was 41%. Considering that the specimens were split into two separate pieces and kept under the same atmospheric pressure sometimes as long as 24 hours, the mending efficiencies are impressive when compared with crack healing of thermoplastic elastomers (12), but they are not as high as those reported for White's composite (4). Another factor, which should affect healing efficiency, is the uniformity and amount of applied pressure. Our preliminary results indicate that the effect of pressure appears to be minimal. The fracture and mending processes of each independently prepared specimen were carried out at least two times. Within experimental error, the critical load at fracture of the third cracking was about 80% of the second, strongly suggesting that this material can be healed efficiently multiple times. The second and third crack usually propagated along the original crack plane. However, it was not uncommon to observe cracking in a region adjacent to a scar from a previous crack. This is probably due to the fact that the healed region has different mechanical properties than the original material and is a result of the unique covalent interfacial bonding.

To test oxidative degradation of the polymer we performed some accelerated tests by heating specimens in air (at standard pressure) to 120° or 130°C and then performing compression testing. Within the experimental error of the test, the mechanical properties showed only minor changes. Though polymeric material 3 has the desirable and unusual properties described above, it has shortcomings as follows that are being improved for next generations: (i) the maleimide monomer melts at too high a temperature, is colored (yellow), and is insoluble in monomer 4F. (ii) Though polymer 3 can be cured in less than an hour at 130°C, a full curing of it takes hours to complete. (iii) The service temperature of polymer 3 is 80° to 120°C, which may be too low for many applications but close to ideal for others, such as self-mending electronic packaging, where cracking occurs due to differences in thermal index of expansion.

  • * To whom correspondence should be addressed. E-mail: wudl{at}


Stay Connected to Science

Navigate This Article