Introduction

Elastomers, a class of polymers renowned for their ability to recover their original shape after deformation, are integral to numerous industries due to their versatile mechanical properties1,2,3,4,5,6,7,8,9,10,11,12. With advancing technology, there is a growing demand for elastomers that offer enhanced mechanical performance to meet increasingly stringent application requirements1,4,6,13,14,15,16,17,18,19,20,21,22,23. Traditional methods of developing these materials through trial and error are not only costly but also inefficient. Consequently, the focus has shifted towards rational design strategies that allow for the precise tailoring of materials to meet specific mechanical demands15,19,20,21,22,24.

In recent years, the role of macromolecular computational simulations has become increasingly indispensable alongside experimental approaches in the field of materials research25,26,27,28,29,30. Techniques such as molecular dynamics have been crucial for predicting the thermodynamic and mechanical properties of macromolecules25,26,27,28, effectively complementing empirical data. Furthermore, advancements in machine learning have significantly sharpened the precision of polymer synthesis and discovery processes29,31,32. Shifting focus to a more fundamental level, small molecule calculations offer a robust and efficient platform for material design due to their inherent simplicity and reduced variability24,33,34. When applied to the optimization and screening of small molecules, these methods have demonstrated their robustness and efficiency19,21. Extending these approaches to predict the behavior of complex macromolecular systems could not only revolutionize polymer design but also refine the design process by offering a streamlined, predictive model that simplifies the intricacies involved in manipulating macromolecular properties. However, a significant challenge remains in translating the insights gained from small molecule calculations into reliable predictions of macromolecular behavior. This necessitates approaches that can capture the complex physical interactions spanning molecular, interchain, and larger structural scales with sufficient accuracy.

In this work, we address these challenges by strategically designing a supramolecular polyurethane (SPU) elastomer system that incorporates supramolecular chain extenders with varied terminal groups. Our findings reveal a significant correlation between the calculated binding energy of supramolecular fragments and key mechanical properties of elastomers, such as tensile strength and toughness. This correlation demonstrates the viability of using small molecule calculations to predict polymer performance effectively (Fig. 1A).

Fig. 1: Conception and system for rational elastomer design.
Fig. 1: Conception and system for rational elastomer design.
Full size image

A Schematic representation of the relationship between small molecule binding energy computations and predictive modeling of polymer mechanical properties. B Representation of an elastomer system incorporating supramolecular chain extenders with diverse terminal groups.

Results

Synthesis of supramolecular elastomers

The SPU elastomers were synthesized by selectively incorporating two of four available aliphatic chain extenders into various combinations: hydrazide-terminated adipic dihydrazide (ADH; H), amino-terminated 1,4-diaminobutane (BDA; A), thiol-terminated 1,4-butanedithiol (BDT; T), and hydroxyl-terminated 1,4-butanediol (BDO; O) as shown in Fig. 1B. Notably, hydrazide, amino, thiol, and hydroxyl groups represent nearly all the typical functional groups that participate in polyurethane chain-extension reactions under mild conditions without interfering with each other. This approach resulted in SPUs with identical hard and soft segments but differing in their combinations of terminal groups from the chain extenders, thereby creating mismatched supramolecular interactions (MMSIs). These elastomers were designated as SPU-HA, SPU-HO, SPU-AO, SPU-HT, SPU-AT, and SPU-OT (Supplementary Figs. 1–6 and Supplementary Table 1). In contrast, elastomers synthesized using individual chain extenders, which form matched supramolecular interactions (MSIs), were labeled SPU-HH, SPU-AA, SPU-TT, and SPU-OO (Supplementary Fig. 7 and Supplementary Table 1). It is crucial to recognize that the consistent presence of identical soft and hard segments within the molecular structures guarantees that all materials prepared in this series are elastomers. This uniformity allows for a precise and thorough examination of the impact that supramolecular fragments have on the system. Proton nuclear magnetic resonance analysis was utilized to confirm the molecular structures (Supplementary Figs. 8–17). Fourier transform infrared (FTIR) spectroscopy provided additional structural insights (Supplementary Fig. 18), and molecular weight distributions were determined via gel permeation chromatography as outlined in Supplementary Table 1. Optical microscopy images (Supplementary Fig. 19) and ultraviolet-visible spectra (Supplementary Fig. 20) confirmed the elastomers’ colorlessness and high transparency.

Quantum chemistry calculations were performed to investigate the binding energies among dimers composed of H, A, T, and O supramolecular fragments (Fig. 2A, Supplementary Fig. 21A, Supplementary Tables 2 and 3, and Supplementary Data). To ensure the calculations mirrored the conditions within the elastomers more closely, ethyl isocyanate was used to cap the terminal groups of the supramolecular extenders. This modification effectively neutralized the impact of free terminal groups on the calculations. Figure 2A depicts the optimized structures of these dimers, showing that their interactions are primarily mediated by hydrogen bonds (H-bonds)35,36 between the imine and carbonyl groups.

Fig. 2: Correlation analysis of binding energy and supramolecular interactions.
Fig. 2: Correlation analysis of binding energy and supramolecular interactions.
Full size image

A Binding modes of the dimers of ADH−ADH (HH), BDA−BDA (AA), BDT−BDT (TT), BDO−BDO (OO), ADH−BDA (HA), ADH–BDO (HO), BDA–BDO (AO), ADH−BDT (HT), BDA−BDT (AT), and BDO−BDT (OT). The gray, white, red, blue, and yellow balls represent carbon, hydrogen, oxygen, nitrogen, and sulfur atoms, respectively. B Binding energies of the dimers plotted against the proportion of H-bonded carbonyl groups in the SPUs. C Binding energies of the dimers plotted against the degree of physical crosslinking in the corresponding SPUs.

FTIR spectroscopy was utilized to quantify the H-bond proportions within the SPUs, with the carbonyl (C=O) absorption bands in the wavenumber range of 1730–1600 cm−1 being deconvoluted into multiple subpeaks (Supplementary Fig. 22). For instance, in SPU-HA, the C=O stretching region was divided into six subpeaks (Supplementary Fig. 22A), indicative of free and H-bonded C=O in urethane, urea, and amide groups. The proportions of H-bonded C=O across the ten elastomers were calculated and are presented in Supplementary Table 4 − 6. These experimentally determined H-bond proportions were correlated with results from density functional theory (DFT) calculations, displaying a strong Pearson correlation (Fig. 2B and Supplementary Fig. 21B). Additionally, the rubber elasticity equation was employed to determine the physical crosslinking degree (Ve) within the SPUs via dynamic mechanical analysis (Supplementary Figs. 23–25). These Ve values were then compared with findings from DFT calculations, revealing a strong Pearson correlation between the two datasets (Fig. 2C). These findings collectively underscore that the binding energy of small molecules aligns with the interactions observed in polymers, with H-bonds being the predominant interactions within the supramolecular fragments.

Correlations between binding energies and mechanical properties

Small-angle X-ray scattering (SAXS) measurements were conducted to analyze the isotropic and micro-phase separation structures of the SPUs (Supplementary Fig. 26). The amorphous nature of the SPUs was further confirmed through X-ray diffraction analyses (Supplementary Fig. 27). Additionally, thermogravimetric analysis was utilized to evaluate the thermal stability of the SPUs, indicating satisfactory performance. These findings are detailed in Supplementary Fig. 28 and Supplementary Table 7.

The mechanical properties of SPU elastomers are exemplified by the stress-strain curves at room temperature under uniaxial extension, the strain rate for all samples is 0.083 s−1 (Fig. 3A, Supplementary Fig. 29). These results demonstrate that elastomers based on MMSIs exhibit enhanced mechanical attributes, including higher strength, stiffness, elongation capability, and tensile toughness. It is worth noting that elastomers with higher H-bond contents (such as SPU-HA and SPU-HH) exhibit a decreasing slope in the large strain region during tensile testing, which can be attributed to the combined effect of strain-induced crystallization and the dissociation of high-density H-bonds. Figure 3B features a comparative bar chart that underscores the significantly enhanced toughness in systems with MMSIs, with a statistical significance analysis corroborating the effectiveness of the MMSI-toughening strategy19. A comprehensive summary of the mechanical and thermal properties of these elastomers is provided in Supplementary Table 7. Among the SPU elastomers, SPU-HA stands out with its high H-bond proportion, large Ve, and strong MMSIs, displaying robust mechanical properties: a tensile strength of 62.0 ± 5.1 MPa and an elongation at break of 3086.7 ± 199.3%, resulting in an ultra-high tensile toughness of 1147.4 ± 160.4 MJ m−3 (Fig. 3A). Notably, SPU-HA not only surpasses the toughness of other transparent and colorless elastomers16,21,22,33,34,37,38,39,40,41 (Fig. 3C) but also exceeds the tensile toughness of Darwin’s bark spider silk, which is renowned for its high toughness at 354 MJ m−342. The true stress at break for SPU-HA reaches 1.9 GPa (Supplementary Fig. 30), comparable to that of typical spider silks (0.8–1.5 GPa)43. Demonstrating its mechanical robustness, a thin-strip sample of SPU-HA weighing 0.42 g successfully lifted a 15.0 kg weight—35,714 times its own weight (Supplementary Movie 1).

Fig. 3: Mechanical properties of supramolecular elastomers.
Fig. 3: Mechanical properties of supramolecular elastomers.
Full size image

A Typical engineering stress-strain curves for elastomers with matched supramolecular interactions (MSIs) and mismatched supramolecular interactions (MMSIs). B Toughness of the elastomers. Statistical significance was assessed using Student’s t-test: ***P < 0.001, **P < 0.01, and *P < 0.05. C Comparison of the ultimate engineering stress and toughness among SPU-HA, other reported colorless and transparent elastomers with toughness >300 MJ m−3, and commercial high-performance elastomers. D Dependence of toughness on the square of tensile strength of the elastomers. E Plot of binding energy for dimers against the tensile strength of the corresponding elastomers. F Binding energy of the dimers plotted in relation to the toughness of the corresponding elastomers. Data are presented as mean ± standard deviation (n = 5). G Strain-dependent dissipated energies of SPU-HA and SPU-AA. H Temperature-dependent loss moduli of SPU-HA and SPU-AA. I Frequency-dependent dielectric loss of SPU-HA and SPU-AA. Scanning electron microscopy (SEM) images of the fracture surfaces of SPU-HA (J) and SPU-AA (K) after tensile testing. L Summary of the properties of SPU-HA and SPU-AA.

With the pronounced micro-phase separation of SPU elastomers, σt versus (λ2 − 1/λ) plots (where σt and λ represent the true stress at break and the elongation ratio, respectively) were used to explore the link between aggregation structure and mechanical performance (Supplementary Fig. 31A, B). The average constraint molecular weight (\(\bar{{M}_{C}}\)) between physical crosslinking junctions was calculated (Supplementary Fig. 31C, D). Notably, the apparently lower \(\bar{{M}_{C}}\) for SPU-HA indicates that mismatched acylsemicarbazide-urea moieties lead to more densely stacked hard domains, facilitating the formation of additional physical crosslinking junctions. This results in shorter distances between junctions, enabling chains to rearrange, orient, and align more effectively during stretching. This structural efficiency endows SPU-HA with high tensile strength and toughness. A moderate correlation between \(\bar{{M}_{C}}\) and the period d value, calculated from the peak of 1D SAXS curves, confirms that the larger \(\bar{{M}_{C}}\) corresponds to a looser packing of hard domains (Supplementary Fig. 32).

The plot in Fig. 3D illustrates a linear relationship between the toughness of the SPUs and the square of their tensile strength, suggesting that H-bonds predominantly govern the failure mechanisms of the elastomers24, thereby enhancing both toughness and strength simultaneously. This analysis substantiates the key role of multiple H-bonds in the strengthening and toughening mechanisms of our SPUs. It is therefore reasonable within our system to use calculated binding energy values, which correlate linearly with H-bond proportion, to predict the mechanical properties of the elastomers. We attempted to correlate binding energy with three indicators of mechanical properties: tensile strength, tensile toughness, and elongation at break. However, when analyzing all elastomers collectively, no consistent relationship was observed (Supplementary Fig. 33). In contrast, separating the MSIs and MMSIs revealed notable Pearson correlations for tensile strength (Fig. 3E), toughness (Fig. 3F), and elongation at break (Supplementary Fig. 34) in relation to binding energies. We observed that the content of H-bonded C=O and Ve are closely linked to the tensile strength of the elastomers (Supplementary Fig. 35). This suggests that the binding ability of the elastomers, reflected by the content of H-bonded C=O and Ve, serves as a critical bridge between the binding energy of small molecules and the mechanical properties of the elastomers. Specifically, the binding energy of small molecules influences the elastomers’ binding ability, which in turn affects their mechanical performance. In assessing material tensile strength, the methodical exclusion of individual data points, except for extremes, and the reanalysis of the remaining dataset resulted in a deviation between the experimental and predicted values of the excluded points, averaging 13.2% (Supplementary Fig. 36). This indicates that the prediction accuracy of our method exceeds 85%, highlighting the potential for using computational analysis of small molecules to predict the performance of complex polymers.

To address potential concerns that the low molecular weight of SPU-OO might affect the reliability of the fitting, we adjusted the reaction conditions to synthesize a high molecular weight SPU-OO sample (Supplementary Figs. 37−40) and incorporated its mechanical properties into the fitting curve. The results still exhibit a clear and consistent trend (Supplementary Fig. 41), further confirming the reliability of the fitting. This also indirectly supports the conclusion shown in Fig. 3D—that the differences in mechanical properties among the elastomers are primarily attributed to variations in secondary interactions.

It is important to note that this correlation may be specific to the systems having supramolecular fragments that share the same middle block but differ in connecting groups, where the predominant type of supramolecular interaction is consistent. Another promising avenue for exploring this correlation is to maintain the same connecting groups while varying the middle block, such as by changing the isocyanates (hard segments such as tolylene-2,4-diisocyanate (TDI), isophorone diisocyanate (IPDI), or 4,4’-diphenylmethane diisocyanate (MDI)). We replaced the diisocyanate (dicyclohexylmethane 4,4’-diisocyanate (HMDI)) with IPDI and successfully synthesized 10 additional elastomers (Supplementary Figs. 42−54 and Supplementary Table 8). The correlation between their mechanical properties and binding energies is shown in Supplementary Fig. 55. The results demonstrate that a strong linear relationship is still maintained, further confirming the reliability and universality of the observed trend. Conversely, our previous work found no linear correlation between calculated and experimental results when both the connecting groups (H-bond interactions) and the middle blocks (involving van der Waals forces, hydrophobic, and π–π interactions) were varied19. When both the connecting groups and the middle segments change, the diversity of supramolecular interactions may increase, introducing additional complexity. For example, the strength distribution of various supramolecular interactions in the polymer system may be uneven, leading to heterogeneous behavior in the material; or the competition or synergy between different supramolecular interactions increases the number of variables affecting the material’s performance, thereby increasing the prediction error. Therefore, this method still has limitations. If it is to be extended to a broader range of supramolecular polymer systems, it will be necessary to consider multiple types of supramolecular interactions in the model and quantify each one. Additionally, the conformation and arrangement of polymer chains, the dynamics of supramolecular interactions, and the overall network structure need to be comprehensively considered.

To further elucidate the difference between MMSI- and MSI-based elastomers, we chose SPU-HA as a representative sample and SPU-AA, containing MSIs, as a control. Both were selected due to their similar binding energies and molecular weights, as well as their robust mechanical properties. We investigated the performances of SPU-HA and SPU-AA under three stimuli: force, heat, and dielectric (Fig. 3G–I and Supplementary Figs. 56−58). Successive loading-unloading tests were conducted, with strain levels increasing from 100% to 1800% (Supplementary Fig. 56), and the dissipated energy was evaluated at different strain levels (Fig. 3G). The results indicated that, due to weak bond fracture at small strains, the energy dissipation of SPU-HA and SPU-AA was not significantly different. However, SPU-HA exhibited significantly higher energy dissipation at larger strains. Temperature-dependent loss modulus (G”) and frequency-dependent dielectric loss analyses revealed that both G” and dielectric loss for SPU-HA were greater than those for SPU-AA (Fig. 3H, I), suggesting that larger internal friction occurs in SPU-HA when subjected to heat or a dielectric environment. We further studied the tensile fracture phenomena of these two elastomers by examining scanning electron microscopy (SEM) images (Fig. 3J, K). The fracture cross-section of SPU-HA exhibited a rough, particle-like morphology, suggesting that its structure was formed by the aggregation of macromolecular chains that slid during the stretching process, indicating significant energy dissipation through crack deflection. In contrast, the fracture cross-section of SPU-AA showed a smooth surface, indicative of extensive damage. The activation energies of SPU-HA and SPU-AA, derived from the stress relaxation behavior, are 87.9 and 89.2 kJ mol−1, respectively, showing only a slight difference, with SPU-HA being slightly lower (Supplementary Fig. 59). Despite SPU-HA exhibiting a significantly higher loss modulus than SPU-AA, their activation energies remain quite similar. This suggests that the stress release process of the material may not be greatly hindered, yet SPU-HA is still able to effectively absorb and dissipate energy under external loading. These findings comprehensively evaluate the effective toughening mechanism of MMSIs and confirm the critical role of interactions between mismatched acylsemicarbazide and urea moieties in enhancing toughness. The mismatched acylsemicarbazide and urea moieties thus provide the elastomer’s molecular structure with enhanced energy dissipation capabilities, enabling it to effectively manage larger loads and strains.

A comparative analysis across 8 dimensions is presented in Fig. 3L. In summary, by integrating MMSI-based acylsemicarbazide-urea-dominant H-bond-reinforced hard domains into the amorphous soft domains, the SPU-HA elastomer demonstrates enhanced performance in terms of energy dissipation and mechanical robustness, which will be elaborated upon in the subsequent section.

Performance of optimized supramolecular elastomers

The microstructural resilience of SPU-HA was evaluated under different applied strains (0%, 200%, 400%, 800%, and release) using in situ wide-angle X-ray scattering (WAXS) and SAXS, as depicted in Fig. 4A–C. Initially, the 2D WAXS patterns (Fig. 4A) display a circular scattering ring, which gradually elongates into an elliptical shape with a prominent principal axis upon increasing tensile strain. Notably, at 400% strain, two distinct diffraction points emerge along the longitudinal axis, corresponding to two peaks in the 1D curve (Fig. 4B), which are ascribed to the crystalline planes of the polytetramethylene ether glycol soft segments. At 800% strain, the scattering ring’s intensity diminishes, and the diffraction points grow more pronounced. Once the strain is released, the scattering ring reverts to its original circular configuration, and the 1D curve returns to its initial state, indicating a reversible microstructural transformation. SAXS analysis also reveals the similar results. The isotropic scattering ring observed in the 2D patterns (Fig. 4A) transforms into a shuttle shape as the tensile deformation progresses, and the corresponding peak on the 1D curve vanishes (Fig. 4C). This change suggests that the original micro-phase separation structure is stretched, leading to a more homogenous internal structure. Upon relaxation of the external force, the scattering pattern returns to isotropic, reaffirming the material’s ability to revert to its original micro-phase separation structure. These observations clearly illustrate the reversible transition from isotropic to anisotropic arrangements within the elastomer during deformation, highlighting its notable microstructural resilience. Figure 4D presents the cyclic tensile curves of SPU-HA subjected to 600% strain across 10 consecutive cycles without delay, followed by an additional cycle after a 3-h rest. The initial loading-unloading curves exhibit a pronounced hysteresis loop, indicative of energy dissipation through the rupture of H-bonds and the intramolecular friction within the polymer chains33. Notably, after a 3-h resting period, the tensile curve largely recovers to its original form, underscoring the robust resilience of SPU-HA. Figure 4E provides a photographic representation of SPU-HA, showcasing its high elasticity.

Fig. 4: Performance of SPU-HA.
Fig. 4: Performance of SPU-HA.
Full size image

A Two-dimensional (2D) wide-angle X-ray scattering (WAXS) and small-angle X-ray scattering (SAXS) patterns under different strains. One-dimensional (1D) WAXS (B) and SAXS (C) profiles under different strains. D Repeated and continuous cyclic stretching with a maximum strain of 600% for 10 times, followed by a 3-h rest period. E Photographic demonstration of the resilience. F Engineering stress comparison at a 200% strain for 1000 cycles. G Typical engineering stress-strain curves of original, healed, and recycled samples. H Comparison of the price-performance ratio between SPU-HA and commercial high-performance elastomers. Inset: Photographic image of the large-scale preparation process using a speed-regulation double-layer glass reactor with a volume of 20 liters. Data are presented as mean ± standard deviation (n = 5).

The fatigue resistance of SPU-HA was further evaluated through 1000 consecutive cycles at a constant 200% strain and a stretching speed of 50 mm min−1 (Supplementary Fig. 60A). Initially, there was a noticeable decline in maximum stress during the stretching phase, primarily due to the partial breakage of H-bonds within the elastomer. However, as the cycles progressed, the maximum stress plateaued, suggesting the establishment of a dynamic equilibrium between the breaking and rapid reformation of H-bonds (Fig. 4F). Additionally, SPU-HA exhibits enhanced puncture, fracture, and impact resistance (Supplementary Figs. 60B–D and 61).

The resistance to ultraviolet (UV) aging was assessed through an aging process that employed alternating cycles of UV light and condensation (Supplementary Figs. 62 and 63). To assess the UV resistance of SPU-HA, we compared it with SPUU-DA,19 an elastomer containing benzene rings that holds the record for tensile toughness among thermoplastic elastomers (Supplementary Fig. 62). Even after 21 days of UV exposure, SPU-HA retains high transmittance in the visible light region, similar to its original state. However, due to the presence of benzene rings, SPUU-DA’s color noticeably darkens and becomes opaque. According to engineers at Q-Lab Corporation, aging a sample for 21 days under these conditions is equivalent to more than a year of exposure in a tropical environment. Thus, this rigorous experimental approach effectively simulates material performance under harsh conditions. Additionally, SPU-HA exhibits reliable temperature-responsive shape-memory properties (Supplementary Figs. 64 and 65), further demonstrating its multifunctional durability and adaptability.

SPU-HA exhibits efficient self-healing properties, allowing it to recover efficiently (Fig. 4G and Supplementary Fig. 66A). Additionally, it can be reprocessed through either hot-pressing or solvent-based methods (Fig. 4G and Supplementary Fig. 66B, C). Temperature-dependent infrared spectroscopy further confirms the reversible dissociation and recombination of H-bonds within the structure (Supplementary Fig. 67 and Supplementary Tables 9−14). These findings emphasize the effective self-healing abilities and recyclability of SPU-HA.

To facilitate larger-scale production, SPU-HA was successfully synthesized in a 20-liter speed-regulated double-layer glass reactor, with no compromise in material performance (Supplementary Fig. 68A and Supplementary Fig. 69). A preliminary evaluation of its industrial scalability was also performed (Supplementary Fig. 68B−D), suggesting promising potential for commercial production. As detailed in Supplementary Tables 15 and 16, the retail price of SPU-HA is set at $6.1 per kilogram. Comparative analysis with five commercially available high-performance elastomers (details can be found in the Methods, these elastomers are randomly anonymized as A, B, C, D, and E for confidentiality and to prevent commercial disputes) shows that SPU-HA offers notable cost-effectiveness, with a performance-to-cost ratio approximately twice that of the leading commercial elastomer (i.e., elastomer E) (Fig. 4H). This competitive pricing provides a significant market advantage, especially for colorless, transparent elastomers. It is poised to transform offerings in sectors such as flexible electronics, impact-resistant materials, and sustainable packaging, where advanced material properties can significantly enhance product performance and sustainability.

Discussion

This study reveals a significant correlation between the binding energy of supramolecular fragments and the mechanical properties of SPU elastomers. This discovery suggests a method that utilizes simple calculations of small molecules to predict the performance of polymer materials. This strategy can direct experimental synthesis efforts, reducing costs and enhancing our understanding of the relationships between supramolecular interactions and material properties. Furthermore, it reduces experimental risks and decreases environmental impact. Crucially, this method effectively bridges the gap between the behaviors of small molecules and the properties of macromolecules. It facilitates the establishment of a more precise theoretical framework, accelerates the identification of potential high-performance materials before synthesis, and enhances our understanding of complex systems. Although further research is necessary to validate its utility across different polymer systems, this pioneering approach represents a significant breakthrough in the field of rational polymer design and sets the stage for future innovations in polymer materials.

Methods

Materials

Polytetramethylene ether glycol (Mn = 2000 g mol–1, PTMEG2000), dicyclohexylmethane 4,4’-diisocyanate (HMDI, purity ≥ 90%), isophorone diisocyanate (IPDI, purity ≥ 99%), 1,4-butanedithiol (BDT, purity ≥ 97%), and 1,4-butanediol (BDO, purity ≥ 99%) were purchased from Aladdin Reagent Company (Shanghai, China). Dibutyltin dilaurate (DBTDL, purity ~ 95%), adipic dihydrazide (ADH, purity ≥ 98%) and 1,4-diaminobutane (BDA, purity ~ 99%) were purchased from Sigma-Aldrich (St. Louis, MO). Five types of commercial elastomers, designated as 101-85D, 2790A, 685A, 65AB, and 1570IC, were purchased respectively from Lubrizol Corporation (Wickliffe, Ohio, United States), Covestro AG (Leverkusen, North Rhine-Westphalia, Germany), BASF (Ludwigshafen, Germany), Huntsman International LLC. (Texas, United States), and Wanhua Chemical Group Co., Ltd. (Yantai, China). All reagents were used as received without further purification.

Synthesis of elastomers

The synthetic route and detailed stoichiometric ratio of reagents for ten elastomers containing ADH, BDA, BDT, and/or BDO were delineated in Supplementary Figs. 1–7 and detailed in Supplementary Table 1. Typically, for SPU-HA synthesis, PTMEG2000 (5.0 mmol) was dried under vacuum at 120 °C for 3 h in a round-bottomed flask, followed by cooling to 80 °C. Subsequently, HMDI (10.0 mmol) and DBTDL (150 µL) dissolved in N,N-dimethylformamide (DMF, 25.0 g) were added into the system, and the mixture was allowed to react for 2 h with thorough stirring to form the prepolymer. Upon cooling to 25 °C, a DMF solution (30.0 g) containing pre-dissolved supramolecular chain extenders (ADH, 2.5 mmol, and BDA, 2.5 mmol) was added to the prepolymer solution under vigorous stirring for 5 h. An appropriate amount of DMF (10.0 g) was added during the reaction to control viscosity, and all steps were conducted under a N2 atmosphere. The obtained viscous liquid was poured into a mold to eliminate bubbles and remove the solvent, and finally, the SPU-HA elastomer was synthesized. According to similar procedures, SPU-HO, SPU-AO, SPU-HT, SPU-AT, SPU-OT, SPU-HH, SPU-AA, SPU-TT, and SPU-OO elastomers were also synthesized, and 1H NMR analyses were shown in Supplementary Figs. 8–17. The synthesis process of IPDI-based elastomers follows a similar procedure, with the only difference being the substitution of HMDI by IPDI as the isocyanate component. A higher molecular weight SPU-OO (HMW-SPU-OO) was obtained by extending the reaction time of SPU-OO fourfold and reducing the solvent amount by half. Furthermore, the number-average molecular weights (Mn) of SPU-HA, SPU-HO, SPU-AO, SPU-HT, SPU-AT, SPU-OT, SPU-HH, SPU-AA, SPU-TT, and SPU-OO elastomers were determined to be ~28.6, 29.8, 46.6, 25.1, 27.5, 29.7, 32.6, 29.3, 40.3, and 12.4 kg mol−1, respectively. The Mn of IPDI-HA, IPDI-HO, IPDI-AO, IPDI-HT, IPDI-AT, IPDI-OT, IPDI-HH, IPDI-AA, IPDI-TT, and IPDI-OO elastomers were determined to be ~48.5, 78.9, 43.6, 25.1, 56.1, 49.2, 54.9, 28.2, 30.4, and 28.6 kg mol−1, respectively. The Mn of HMW-SPU-OO was determined to be ~28.8 kg mol−1. Unless otherwise specified, each synthetic experiment of the elastomers was repeated at least three times. To ensure scientific comparability, the synthesis parameters of the elastomers should be kept as consistent as possible.

Healing for elastomers

The healing process involved cutting an 80% notch in the middle of the sample film and then placing it in an 80 °C oven for 6 h with trace amounts of DMF.

Characterizations

The molecular weight and dispersity of elastomers were measured using a Waters 1525 gel permeation chromatography (GPC) instrument. Proton nuclear magnetic resonance (1H NMR) analyses were conducted on a Bruker AVANCE III HD instrument (400 MHz) with DMSO-d6 as the solvent. Fourier transform infrared (FTIR) spectroscopy was performed on a Bruker Tensor II spectrometer in the attenuated total reflection (ATR) mode at 25 °C. The thermal stabilities of elastomers were assessed utilizing a thermogravimetric analyzer (TGA 5500, TA Instruments). X-ray diffraction (XRD) analyses were performed on a SmartLab 9KW X-ray diffractometer with Cu Kα radiation (λ = 0.154 nm), and the scanning speed was 20° min−1. Transparency tests were conducted using an ultraviolet-visible (UV-vis) spectrophotometer (Cary 5000), scanning wavelengths from 200 to 800 nm with air as the reference. Shore hardness values were determined with a Shore hardness tester (Nscing Es LX-A). The dielectric properties of elastomers were studied on an Agilent 4294A precision impedance analyzer. Rheology measurements were carried out on an Anton Paar MCR302 instrument, employing the 25 mm parallel plate geometry. The pendulum impact test was performed on a JJ TEST XJJD 50 instrument, where the nominal energy of the pendulum bob was set at 50.0 J, and the impact velocity at 3.8 m s–1. Impact strength values were calculated using the software included with the instrument. Temperature-dependent infrared spectra were recorded on a Nicolet 6700 spectrometer equipped with a PIKE gold-plated integrating sphere, scanning the sample at different temperatures across a wavelength range of 4000–600 cm−1. The spectra were collected in the temperature range of 30–140 °C, with a heating rate of 5 °C min−1 and measurements taken at 10 °C intervals. Scanning electron microscopy (SEM) images were recorded using a Hitachi SU8010 scanning electron microscope under vacuum. The samples were coated with a thin layer of platinum (2–3 nm) before imaging. The UV resistance testing of elastomers was carried out using an accelerated weathering tester (QUV/spray). The test employed UV light at a wavelength of 340 nm with a light intensity of 0.77 W m−2 at 60 °C for 8 h, followed by a condensation stage at 50 °C for 4 h. This alternating cycle was maintained throughout the testing period.

A Bruker NANOSTAR device was utilized to acquire two-dimensional (2D) small-angle X-ray scattering (SAXS) patterns. Each 2D SAXS experiment involved exposing the sample to the X-ray source for 10 min. The X-ray radiation had a wavelength of 0.154 nm, and the sample-to-detector distance was set at 1060 mm. The one-dimensional (1D) SAXS curves were generated using the DIFFRAC.SAXS software. The periodicity (d) was calculated according to the Bragg’s Law:

$$d=\frac{2\pi }{{q}_{\max }}$$
(1)

where qmax corresponds to the peak position of the 1D SAXS curve.

Measurements of 2D wide-angle X-ray scattering (WAXS) were carried out using a Bruker NANOSTAR device. The system was furnished with a Cu Kα X-ray generator, and the X-ray radiation had a wavelength of 0.154 nm. Each WAXS pattern, acquired at the center of the sample, was collected over a 300 s duration, with a sample-to-detector distance set at 59 mm. Standard background correction procedures were applied to all WAXS patterns. The 1D WAXS profiles were then generated using the DIFFRAC.SAXS software.

All mechanical tensile experiments were carried out on an Instron 3344 universal testing machine (500 N). Unless otherwise specified, the testing rate was 50 mm min–1 under ambient conditions. At least five individual tensile tests were measured for each kind of sample. The strain rate of 0.083 s−1 is derived by dividing the stretching speed (50 mm min−1) by the sample gauge length (10.0 mm). Young’s moduli were determined by the slope of the engineering stress-strain curve in the initial linear region within a strain of 2%. The tensile toughness, defined as the area under the engineering stress (σ)–strain (ε) curve, was calculated using the following equation:

$${{{\mathrm{Toughness}}}}={\int }_{0}^{{{{{\rm{\varepsilon }}}}}_{\max }}{{{\rm{\sigma }}}}\, d{{{\rm{\varepsilon }}}}$$
(2)

True stress (σt) and true strain (εt) were calculated based on the engineering stress-strain curves, according to the following equations:

$${\sigma }_{t}=\sigma \left(\varepsilon+1\right)$$
(3)
$${\varepsilon }_{t}={{\mathrm{ln}}}\left(\varepsilon+1\right)$$
(4)

Cyclic tensile tests were performed on elastomers with identical dimensions, employing the maximum strain of either 600% or 200%. Dissipated energy was calculated by integrating the area encompassed by the cyclic tensile curves. The hysteresis ratio was calculated by the ratio of dissipated energy to the input energy. Puncture resistance tests were carried out using a sample-holding apparatus and a custom-made metal needle with a tip radius of 450 µm.

The fracture energy was measured using a sample featuring a notched edge at a stretching rate of 3 mm min–1. Tests were conducted on specimens both with and without a notch, with gauge dimensions of 25.0 mm in length, 5.0 mm in width, and 0.5 mm in thickness. The fracture energy (Gc) was calculated using the following equation:

$${G}_{c}=\frac{6{Wc}}{\sqrt{{\lambda }_{c}}}$$
(5)

where c is the length of the notch (1.0 mm), λc represents the elongation-at-break of the notched sample, and W is the strain energy calculated by integrating the stress-strain curve of the unnotched specimen until εc (εc = λc – 1).

The storage modulus, loss modulus, and tan δ as functions of temperature were measured via dynamic mechanical analysis (DMA) on a Metravib DMA 50 instrument, with the tensile mode. Rectangular geometry samples (25.0 mm × 10.0 mm × 0.5 mm) were measured from −100 °C to 100 °C, at a heating rate of 3 °C min−1 and a constant frequency of 1 Hz. The physical crosslinking degree (Ve) of elastomers was qualitatively calculated using the rubber elasticity equation:

$${E}^{\prime}=3{{{\mathrm{RTV}}}}_{e}$$
(6)

where the rubbery platform modulus (E’) is the storage modulus at the temperature of T (40 °C above Tg), and R is the universal gas constant. The crosslinking densities of the crosslinked networks closely relate to their rubbery platform moduli.

A quantitative shape memory experiment was conducted on a TA instrument Discovery 850 system under strain-controlled conditions with a film-tension clamp. Rectangular specimens of ~25 mm × 5 mm × 0.5 mm were employed for the test. The shape fixation rate (Rf) and shape recovery rate (Rr) of the sample were calculated by the following formula:

$${R}_{f}=\frac{{\varepsilon }_{3}-{\varepsilon }_{1}}{{\varepsilon }_{2}-{\varepsilon }_{1}}\times 100\%$$
(7)
$${R}_{r}=\frac{{\varepsilon }_{3}-{\varepsilon }_{4}}{{\varepsilon }_{3}-{\varepsilon }_{1}}\times 100\%$$
(8)

where ε1 is the initial strain before deforming, ε2 is the maximum strain of the sample under stress, ε3 is the strain after stress removal, and ε4 is the recovered strain at recovery temperature.

A TA instrument Discovery 850 system was used in the stress relaxation experiments, and the samples were subjected to a strain of 5% at various temperatures. By the Maxwell viscoelastic fluid model, the relaxation time (τ) was determined as the time taken for the sample to relax to 1/e of the initial stress. The correlation between the relaxation time (τ) and temperature adhered to the Arrhenius equation:

$${{{\mathrm{ln}}}}\, \tau={{{\mathrm{ln}}}}\, {\tau }_{0}+\frac{{E}_{a}}{{RT}}$$
(9)

where τ is the relaxation time, τ0 is the characteristic relaxation time at infinite temperature, T is the testing temperature, and R is the universal gas constant.

Calculation method for \(\overline{{{{{\boldsymbol{M}}}}}_{{{{\boldsymbol{C}}}}}}\)

The details are as follows: σt versus (λ2 − 1/λ) curves are derived from σt versus strain (ε) curves. In this context, λ, denoting the elongation ratio, is defined as l/l0, where ε = l/l0 − 1 = λ − 1, with l representing the length of the sample after drawing and l0 being the gauge length. The curve σt versus (λ2 − 1/λ) can be fitted using σt = Y + Gp (λ2 − 1/λ), where Y is related to the extrapolated yield stress, and Gp is the strain hardening modulus. Furthermore, Gp can be expressed as a function of ρRT/\(\bar{{M}_{C}}\), where the densities of materials (ρ) are measured at ~0.74 g cm−3 for SPU-HA and 0.95 g cm−3 for SPU-HO, R (J mol−1 K−1) is the ideal gas constant, T (K) is the absolute temperature, and \(\bar{{M}_{C}}\) (g mol−1) is the average constraint molecular weight between the physical crosslinking junctions.

Methods of predictive presentation

The accuracy of the prediction was measured using the prediction deviation value (γ). The γ for the tensile strength of the sample was determined using the equation:

$$\gamma=\left|\frac{{\sigma }_{{{\mathrm{actual}}}}-{\sigma }_{{{\mathrm{theoretical}}}}}{{\sigma }_{{{\mathrm{theoretical}}}}}\right|\times 100\%$$
(10)

where σactual is the actual tensile strength, and σtheoretical is the predictive tensile strength according the prediction curve using the remaining points. For the prediction curves, we generated each by systematically excluding individual data points and fitting a regression line to the remainder of the dataset. The identification of maximum and minimum binding energy values not only encompasses the overall data range but also underscores the relevance and robustness of the fitted curve, justifying our decision to retain these extremes.

Noda’s rule for the generalized 2D correlation spectra

If the correlation intensity Ф (ν1, ν2) in synchronous spectra shares the same sign (either “+” or “−”) as the correlation peak Ψ (ν1, ν2) in asynchronous spectra, it indicates that the movement of band ν1 precedes that of ν2, and vice versa. Additionally, if the correlation in synchronous spectra is nonzero, yet zero in asynchronous spectra, then the movements of bands at ν1 and ν2 occur simultaneously.

The Noda’s rules are summarized as follows: (1) If Ф (ν1, ν2) > 0, Ψ (ν1, ν2) > 0 or Ф (ν1, ν2) < 0, Ψ (ν1, ν2) < 0, then the movement of ν1 occurs prior to that of ν2. (2) If Ф (ν1, ν2) > 0, Ψ (ν1, ν2) < 0 or Ф (ν1, ν2) < 0, Ψ (ν1, ν2) > 0, then the movement of ν1 occurs after that of ν2. (3) If Ф (ν1, ν2) > 0, Ψ (ν1, ν2) = 0 or Ф (ν1, ν2) < 0, Ψ (ν1, ν2) = 0, then the movements of ν1 and ν2 occur simultaneously. Ф (ν1, ν2) and Ψ (ν1, ν2) represent the correlation peaks in the synchronous and asynchronous spectra, respectively.

Computational details

All-electron density functional theory (DFT) calculations were performed using the ORCA quantum chemistry software44 (Version 5.0.3). Geometry optimization and frequency analysis of the dimers were carried out with the B3LYP/def2-SVP method45,46, applying a frequency scaling factor of 0.9640 for accurate frequency corrections. Singlet point energy calculations were conducted with a larger basis set, def2-TZVP. To account for weak interactions, the DFT-D3 dispersion correction with Becke–Johnson (BJ) damping47,48 was employed. The geometrical Counterpoise (gCP) correction49 was used to eliminate artificial over-binding effects arising from basis set superposition error (BSSE). The binding energy (∆E) between X and Y was calculated using the following formula:

$$\Delta E={E}_{{XY}}-\left({E}_{X}+{E}_{Y}\right)$$
(11)

The noncovalent interaction nature was investigated using the Independent Gradient Model (IGM) method through Multiwfn software50. Visualization of IGM and orbitals was rendered using Visual Molecular Dynamics (VMD)51. The atomic coordinates of optimized computational models are available in the Supplementary Data.

To more closely approximate the actual system, ethyl isocyanate was used to cap the terminal groups of the supramolecular extenders, resulting in the formation of four distinct supramolecular fragments: ADH (H), BDA (A), BDT (T), and BDO (O).

Analysis of the economic efficiency of SPU-HA

Assuming that the process of scale-up involves four 1000-L reactors operating stably for 10 consecutive years, with a 10% loss of DMF per cycle, a net production capacity of 10 kg h−1, an electricity price of $0.097 per kilowatt-hour, a water price of $0.581 per ton, a consumption of $0.277 per hour of ethyl acetate, and without considering factors such as labor costs, taxes, equipment depreciation, and marketing expenses, we can calculate the unit cost of SPU-HA. It should be noted that this section is only a rough estimate, aiming to demonstrate the scalable preparation and price advantage of SPU-HA.

The hourly cost of raw materials based on their consumption rates and market prices is shown in Supplementary Table 15. By incorporating indirect manufacturing costs, including equipment fees and electricity expenses, the total cost of continuous production over a decade can be ascertained (Supplementary Table 16).

The unit cost is determined using the following formula:

$${{{\rm{Unit\; cost}}}}=\frac{({{{\rm{Material\; cost}}}}+{{{\rm{Equipment\; cost}}}}+{{{\rm{Electricity\; cost}}}})}{{{{\rm{Total\; output}}}}}$$
(12)

where the total output over a decade is 876 tons. The equipment costs total $0.573 million, and electricity costs, based on a rate of $0.097 per kilowatt-hour and an hourly power consumption of 146 kilowatts, amount to $1.239 million. The direct material cost is $1.508 million. Therefore, the calculated unit cost of SPU-HA is $3.8/kg.

Statistical analysis

Statistical analysis was conducted using the Origin software. Data are presented as means ± standard deviation (SD, n = 5) unless otherwise stated. Statistical significance was assessed using Student’s t-test: Linear regression model was fit via ordinary least square in the correlation studies. The Pearson correlation coefficient (r) and the associated P values were calculated to examine the correlations. The level of significance in all statistical analyses was set at a probability of P < 0.05 (one-tailed).