Preparing a strong self-healing elastomer that functions at ambient temperature
The self-healing TPU referred to as C-IP-SS was synthesized from poly(hexamethylene carbonate) diol (C, a carbonate-type aliphatic macrodiol) as the soft segment, asymmetric alicyclic isophorone diisocyanate (IP) as the hard segment, and an aromatic disulfide (SS) as the chain extender (Fig. 1a and Supplementary Fig. 1)39–44. SS metathesis drives intrinsic self-healing at room temperature45,46. Our previous self-healing TPU (referred to as E-IP-SS), which contains ether-type macrodiol (E), IP, and SS units47, was used as the control. Both self-healing TPUs have equal (15 wt%) macrodiol compositions and number-average molecular weights (Mns) of 1 kg mol− 1 (Supplementary Figs. 2–6 and Supplementary Table 1). A commercial TPU (Es-MD) containing ester-type macrodiol (Es) and methylene diphenyl diisocyanate (MD) units was used as a non-self-healable control. The weight-average molecular weights (Mws) of C-IP-SS, E-IP-SS, and Es-MD were measured to be 24, 46, and 99 kg mol− 1, respectively.
Interestingly, C-IP-SS exhibited significantly superior mechanical properties compared to other self-healing materials. The tensile properties of the three types of TPU film prepared by solvent casting were examined. Prior to this study, E-IP-SS exhibited the highest toughness (26.9 MJ m− 3) and UTS (6.8 MPa) among self-healing materials that function at ambient temperature (Fig. 1b). Nevertheless, its tensile performance is trivial compared to that of Es-MD, with a toughness and UTS of 115 MJ m− 3 and 36 MPa, respectively. Although C-IP-SS has the lowest Mw among the TPUs, it has the highest UTS of 43 MPa and exhibits a 2.8-fold greater toughness (75 MJ m− 3) than E-IP-SS (Fig. 1c).
The outstanding tensile properties of C-IP-SS have little penalty on its self-healing efficiency. C-IP-SS films were cut in half, reattached, and then allowed to recover at 35 °C for varying amounts of time (Fig. 1b, c and Supplementary Table 2). UTS recoveries of 29, 39, 65, and 77% were observed after 1, 6, 24, and 48 h, respectively. Notably, the UTS of the recovered C-IP-SS exceeded that of virgin E-IP-SS after only 1 h of self-healing; after 48 h, the recovered C-IP-SS is as strong as Es-MD (Fig. 1c, d and Supplementary Movie 1), and another cut and re-spliced film with a contact cut-area of 30 mm × 3 mm was able to readily lift a 10-kg weight (Fig. 1e and Supplementary Movie 2). Even after 1 min of recovery at room temperature (25 °C), a cut and re-spliced film with a contacted cut-area of 5 mm × 1 mm was able to withstand manual drawing and twisting (Fig. 1f and Supplementary Movie 3). According to the Ashby plot of UTS versus self-healing temperature for recently reported elastomers, previous tensile strengths did not exceed 20 MPa without the aid of light or a temperature above 40 °C (Supplementary Fig. 7 and Supplementary Table 3)9,21–24,34−40,42,43,47.
Achieving Both Self-healing And Strength Through Reversible Structural Change
The breakthrough in the trade-off between self-healing and strength achieved for C-IP-SS is attributable to a dual mode of operation involving its mechano-responsive H-bonding array, with reversible strain-induced crystallization providing a major contribution. In the stress–strain curve for C-IP-SS (Fig. 1b), the concave upward trend is a clear sign of strain hardening. Figure 2a reveals that the tensile moduli (i.e., differential stress values) increase sharply with strain and are substantially different to those of E-IP-SS and Es-MD. C-IP-SS exhibits a more than 4.3-fold higher modulus than the other TPUs at 400% elongation, and becomes translucent due to the formation of crystallites, while the other TPUs remain relatively transparent (Fig. 2b, c and Supplementary Figs. 8 and 9). Interestingly, the original strain and transparency of C-IP-SS are fully recovered upon unloading (Supplementary Fig. 10 and Supplementary Movie 4); strain-induced crystallization upon loading is superficially incompatible with full recovery upon unloading; it cannot be explained either by the solid crystallization of semi-crystalline polymers or entropic elastomer elasticity because the former does not facilitated strain recovery, while the latter does not enable phase transition. Inspired by the recovery process of biomaterials, we deduce that rapid switching between the two individual internal structures drives the reversible deformation of C-IP-SS (Fig. 2d)48.
The unusual strain-induced crystallization of C-IP-SS is strongly correlated with the reversible disorder-to-order switching of the mechano-responsive H-bonding array. Therefore, we separately investigated the internal H-bonding structures of C-IP-SS in its static and dynamic states by Fourier-transform infrared (FT-IR) spectroscopy (Fig. 3). As evidenced by differential scanning calorimetry (DSC), C-IP-SS is poorly microphase separated, with non-crystalline hard segments in the static state because the non-planar SS and IP moieties interfere with hard-segment stacking (Supplementary Fig. 11)49. This behavior is unusual because typical TPUs have distinctly separated amorphous soft-domain and crystalline hard-domain microphases induced by the intermolecular stacking of mesogenic MD. The FT-IR spectrum of Es-MD shows an intense band at 1703 cm− 1 that corresponds to H-bonded urethane groups in the hard domain (Supplementary Fig. 12), and conventional carbonate-type TPUs that contain MD also exhibit two dominant bands at approximately 1736 and 1699 cm− 1 from their respective soft and hard domains41,50. On the contrary, the FT-IR spectrum of C-IP-SS shows four different bands associated with carbonyl (C = O) groups that are assigned to ester groups that neighbor SS units (1645 cm− 1), those that are not H-bonded or are in disordered soft segments (1741 cm− 1), are involved in H-bonding between hard and soft segments (1718 cm− 1), and H-bonded between hard segments (1692 cm− 1) (Fig. 3a)51,52. As a result, the band at 1718 cm− 1 due to H-bonding between hard and soft segments is as strong as the band at 1692 cm− 1 associated with H-bonding between hard segments. The –NH– units of each hard segment could form H-bonds with abundant soft segmental carbonyl groups53. Therefore, in static mode, C-IP-SS has a relatively high degree of disorder with regard to its non-crystalline hard segments and the high concentration of H-bond-free carbonyl groups.
Uniaxial stretching in the dynamic state enhances the degree of H-bond ordering, leading to an increase in crystallinity and a decrease in the concentration of carbonyl groups that are not H-bonded, which is well-represented by a 2D gradient FT-IR map, in which the first derivative of the absorbance (A) as a function of the extension % (E) (i.e., dA/dE) is constructed in E versus wavenumber space (Fig. 3b)54. The red and gray contours indicate positive and negative dA/dE, respectively. The strongest red and gray contour lines at 1718 and 1741 cm− 1, respectively, confirm that extension strengthens the adsorption band corresponding to H-bonding between hard and soft segments, but weakens the band associated with the absence of H-bonding or disordered soft segment, which is ascribable to the disorder-to-order transition of H-bonds between the laterally aligned stretched carbonate chains (i.e., crystallites) (Fig. 3c). As expected, the adsorption-band pattern fully returns to that of the original state after release. Thus, the mechano-responsive H-bonding structure undergoes array switching.
Rheology was used to explore whether or not the internal H-bonded structure affects the mechanics of C-IP-SS in static mode. Oscillating rheological testing at very low strain (1%) is a useful tool for investigating the static mode because it measures linear viscoelastic behavior. Each TPU was subjected to frequency sweep testing at 25 °C (Fig. 4a and Supplementary Fig. 13) to provide its complex viscosity at 0.05 rad s− 1 (η*0.05), yield stress, and the slope of the Cole − Cole plot. In short, higher values of η*0.05 and yield stress, and lower Cole − Cole-plot slopes indicate that C-IP-SS is a more solid-like system20,55.
The temperature-dependent loss tangents (tan δ) for the three TPUs were investigated in the 25–95 °C range (Fig. 4b). If tan δ shows a decrease (or increase) by a standard value of unity, then the system exhibits solid-like (or liquid-like) behavior56. In contrast to the tan δ value of E-IP-SS, that of C-IP-SS remained below unity over the entire examined temperature range and was lowest near room temperature, which is a meaningful result because it reveals that a product made of C-IP-SS will be relatively safe from melting at high temperature. Moreover, C-IP-SS exhibited the highest Young’s modulus (elastic modulus at a low tensile strain) of 15.5 MPa which is 11-, and 1.7-fold higher than those of E-IP-SS, and Es-MD, respectively (Supplementary Table 2). These data suggest that C-IP-SS is the most solid-like TPU among those studies owing to its rich carbonyl H-bonds, even though it has a four-fold lower Mw than Es-MD. As revealed by FT-IR spectroscopy, H-bonds are created between hard segments of C-IP-SS and all other hard and soft segments to balance the disordered chains, while the H-bonds in Es-MD are relatively fixed between ordered hard segments, but are lacking between soft segments or disordered chains. Consequently, the former system exhibits a high degree of disorder while constructing a structure with a higher density of internal H-bonds than the latter.
As a motif for self-healing, the molecular process in static mode can be described by rheological master curves of storage (G') and loss (G") moduli (Fig. 4c). A typical viscoelastic polymer shows four zones that are characterized by trends in G' and G" (Fig. 4d) that are ordered as terminal, rubbery plateau, transition, and glassy zones57. Comparing the two self-healable TPUs reveals that E-IP-SS only exhibits the first two zones, with a clear terminal region (G" > G'). On the other hand, C-IP-SS shows three zones, from terminal to transition, with a dominant rubbery plateau region (G' > G"). These observations suggest that the chains in C-IP-SS experience a lower degree of molecular slips than those in E-IP-SS. As a quantitative parameter, the flow transition relaxation time (τf) was determined from the reciprocal frequency at the G'/G" crossover between the terminal and rubbery zones, which was found to be > 2 years for C-IP-SS and 112 s for E-IP-SS (Fig. 4e), suggesting that C-IP-SS behaves more like a glassy solid than E-IP-SS; hence, chain-flow relaxation barely contributes to C-IP-SS healing on a reasonable time scale.
The key to C-IP-SS self-healing can be revealed from the segmental relaxation time (τs), which was calculated from its relaxation time curve at an angular frequency (ω) of 0.05 rad s− 1 (Supplementary Fig. 14). The τs of C-IP-SS was calculated to be 8.5 s, very similar to that of E-IP-SS (7.4 s) (Fig. 4e). C-IP-SS self-healing is therefore a consequence of high segmental vibrations that enable the exchange of disordered H-bonds, and dynamic SS in the poorly microphase-separated system. We also note that classical chain-flow-meditated self-healing degrades mechanical properties, while segmental motion-driven self-healing minimizes the mechanical-property penalty. The difference between the molecular dynamics of flow transition and segmental motion is schematically illustrated in Supplementary Fig. 15.
X-ray tools can be used to trace the structural changes that occur as TPUs transition from static to dynamic mode during stretching (Fig. 5a, b). The long- and short-range structural regularities of a TPU can be examined by small- and wide-angle X-ray scattering (SAXS and WAXS) techniques, respectively. To investigate static mode, we examined undrawn samples. While Es-MD presents characteristic SAXS patterns arising from clear phase separation (Supplementary Figs. 16 and 17), both C-IP-SS and E-IP-SS do not exhibit distinct SAXS patterns because they are poorly phase separated. When examined by WAXS, only Es-MD displayed a dim peak, which is evidence for the presence of crystalline hard segments (Fig. 5a and Supplementary Fig. 18).
To investigate dynamic mode, we analyzed strained translucent specimens. C-IP-SS and E-IP-SS still failed to show any SAXS peaks, indicating that external stress does not lead to phase separation. C-IP-SS showed the most noticeable change in its 1D WAXS spectrum when deformed at high strain (Fig. 5b), which suggests that C-IP-SS undergoes a distinct structural conversion into a crystalline form. Two peaks appeared at q values of 14.0 and 16.1 nm− 1 in the 1D WAXS spectrum when stretched to 400%; these peaks correspond to the (020) and (110) planes of crystalline methylene groups, respectively58, with two clear spots observed in the equatorial region of the 2D WAXS pattern (Fig. 5a). Stretched carbonate-type soft segments have relatively regular and aligned secondary structures and therefore form metastable crystals in which methylene segments are laterally packed with the aid of the ordered H-bonding array (Supplementary Fig. 19). Consequently, C-IP-SS stiffens and becomes as strong as commercial Es-MD. On the other hand, E-IP-SS shows a strain-independent amorphous halo 2D WAXS pattern, and 400% elongation of Es-MD provides a clearer spot in the equatorial region of its 2D WAXS pattern and a small change in the 1D pattern compared to that prior to stretching. The hard domain is oriented along the stretching axis with minor crystallite growth. The minor amounts of strain-induced crystallization reveals that Es-MD and E-IP-SS behave like classic elastomers that observe almost isothermal entropic elasticity (i.e. ∆H or ∆U ≈ 0).
The 2D pattern of the relaxed C-IP-SS was acquired (Supplementary Fig. 20) and reveals that its metastable crystals had fully returned to the original amorphous phase upon unloading due to the instantaneous nature of its ordered H-bonding array. Meanwhile, the strain recovery of C-IP-SS degraded when held for a prolonged time (72 h) at 400% strain (Fig. 5a, b), which indicates that the metastable crystal structure is stabilized upon long-term extension due to more organized lateral stacking and H-bonding.
Practical Calculation Of Reversible Mechano-responsive Crystallization
The full strain recovery of C-IP-SS is consistent with the phase-conversion-induced elasticity observed for biomaterials59. The egg-capsule wall of a channeled whelk was reported to exhibit highly reversible extensibility in which internal-energy return prevails over entropic relaxation upon unloading60. In other words, the reversible deformation of C-IP-SS is a non-isothermal process with an internal energy change that occurs through mechano-responsive crystallization. The internal energy contribution to the elasticity of C-IP-SS was evaluated on the basis of a well-established thermodynamic framework, as encapsulated by the following Eq. (1):
where U and S are the internal energy and entropy of the system, respectively, and L, T, and V are the length, temperature, and volume of the specimen, respectively. The total elastic force (f) is the sum of the changes in U and S as functions of L (fU and fS) at constant V and T48; fS is determined from an f versus T term through Maxwell’s relationship. Figure 5c displays tensile stress–strain graphs of C-IP-SS at various temperatures between − 30 and 40 °C, from which f versus T plots at particular strains (i.e., L) were constructed (Fig. 5d). A classic elastomer has a positive slope because it is an isothermal system; hence fU = 0 and S decreases with L, and fS < 0, which leads to entropy-driven elasticity29,61. However, the plots for C-IP-SS have negative slopes at all examined L and T values, which means that stretching changes the internal energy (i.e. fU ≠ 0). The U term prevails over the S term for f. It is quite noticeable that the slope critically changes at 0 °C, which is close to the glass-transition temperature (Tg) (Supplementary Fig. 21). The steeper slope indicates that a higher internal-energy change results in more-stable and ordered crystals, which is associated with the following observations. C-IP-SS was drawn and released above its Tg; the original strain is then recovered because of its relatively unstable crystal phases. However, C-IP-SS is not reversible below Tg because relatively stable crystals are formed. The energy (ΔU) required to form the stress-induced metastable crystal is calculated to ~ 6.47 cal cm− 3 above Tg, the details of which are described in Supplementary Note 1 (Supplementary Figs. 22–24 and Supplementary Table 4). At a TPU density of 1.25 g cm− 3, this value can be converted into ~ 5.17 cal g− 1. The calculated heat of the metastable crystal is 3 − 10-times lower than those (15 − 65 cal g− 1) of stable commercial-polymer crystals62, which qualitatively shows that the instantly formed strain-induced crystal is unstable.
The toughness, resilience, and self-repairability features of C-IP-SS resemble those of biological tissue. In particular, we expect that this TPU can be used as a precursor material for artificial muscle due to its high mechanical and elastic-recovery performance. Accordingly, the in vivo biocompatibility of C-IP-SS was evaluated by the Daegu Gyeongbuk Medical Innovation Foundation (DGMIF), a contract clinical research organization, following the ISO 10993-6 Annex A standard63, and ethically approved by the Institutional Animal Care and Use Committee (Korea) (IACUC; approval code DGMIF-19042402-00). Experimental sample films of C-IP-SS (diameter: 10 mm) were implanted in the subcutaneous tissue of rats (n = 5), and a polyurethane film containing residual DMSO (solvent) and a high density polyethylene (HDPE) film were used as positive and negative controls, respectively (Fig. 6a). The rats were euthanized after 12 weeks and the subcutaneous tissue samples were histopathologically analyzed after routine fixing and dyeing. The histopathological tissue images reveal that C-IP-SS produced less inflammatory cells than the positive control (Fig. 6b). Inflammatory responses were scored semi-quantitatively by a pathologist following the ISO 10993-6 inflammatory-reaction intensity guidelines: non-irritant < slight < moderate < severe. The C-IP-SS film scored the lowest inflammatory intensity; i.e., non-irritant. In the preliminary experiments performed in this study, C-IP-SS showed superior biocompatibility and did not elicit chronic or severe inflammation.