Ordered nanoplastic-elastomer networks resolve conflict between softness and stability Authors: Yan Wang, Zhangkan Lin, Zheqi Chen, Guodong Nian, Shaoxing Qu, Yingwu Luo Published: 2026 ABSTRACT AbstractSoft materials often fail through snap-through instability, where a small increase in load causes a sudden, catastrophic deformation. However, overcoming this instability requires a polymer network of two seemingly contradictory behaviors: softness at small strains to allow deformation, but early stiffening at afterward strains to prevent instability. Here we resolve this conflict by designing an architecture of ordered nanoplastic-elastomer network. We identify two design principles: a small volume fraction of rigid plastic nanodomains is orderly arranged within a soft elastomer matrix; the nanodomains and matrix are strongly linked by covalent bonds. These features together produce a crucial effect: macroscale strain is greatly amplified at the microscale, inducing earlier stiffening while retaining small-strain softness. Theoretically and experimentally, we demonstrate that this network architecture can prevent notorious premature failure in dielectric elastomer actuators, and greatly enhance the actuation performance. These results suggest a general route to design soft materials that resist catastrophic instability-induced failure.Subject terms: Polymers, Mechanical engineering, Organic molecules in materials science Introduction Soft materials capable of large deformation are indispensable in both nature and engineering, yet they share a common vulnerability: mechanical instabilities1. Under a compression or biaxial stretch, a thin rubber sheet1, a balloon2, a dielectric elastomer actuator3, or even a blood vessel wall4 can all undergo sudden snap-through deformation when pushed beyond a critical load. In this process, the system abruptly jumps between two distinct deformed states. Such instabilities can be harnessed for function, as in plant seed dispersal or engineered switches5–8, but in many cases they are catastrophic, leading to cavitation1,9, premature electrical breakdown3,10, or tissue rupture3,10. The difficulty in resisting instability arises from the inherent mechanics of soft polymer networks, which remain soft until their strands approach full extension. Such delayed stiffening allows instabilities to grow uncontrollably before the material can mount sufficient resistance. Scaling laws in polymer physics capture this dilemma11–14. Consider a polymer network with n of repeat units on a network strand. Modulus scales inversely with strand length, ~n−1, whereas the stretch at which stiffening sets in scales positively with chain length, ~n1/2. Thus, the networks that are soft enough for large deformations stiffen too late to prevent instability, while the networks that stiffen early are too rigid and brittle to be useful. This conflict between softness and stability limits the design space for actuators, seals, and even biological mimics. Addressing this dilemma requires two seemingly contradictory properties: softness to ease deformations and early stiffening to suppress instability—demands that conventional network topologies cannot meet. In this paper, we resolve the conflict between softness and stability by designing an architecture of Ordered nanoPlastic-Elastomer Network (OPEN). In this design, rigid plastic nanodomains are orderly arranged within a soft elastomer matrix, and dense covalent bonds link the domains and the matrix (Fig. 1a). Upon a stretch, the rigid nanodomains do not deform, and the dense covalent interlinks transmit stress. The architecture induces a crucial effect of strain amplification: by incorporating a small volume fraction of the rigid domains, macroscopic stretch is greatly magnified within the soft matrix, triggering early strain stiffening without compromising softness at small strains (Fig. 1b). By both theory and experiment, we demonstrate that incorporating 5 vol.% of rigid nanodomains eliminates snap-through instability and prevents premature failure in dielectric elastomer actuators. Furthermore, theoretical analysis indicates that the same architecture can also suppress pneumatic snap-through in inflated elastomer membranes. Ordered nanoplastic-elastomer network amplifies local strain. a In an ordered nanoplastic-elastomer network, rigid plastic nanodomains are orderly arranged within a soft elastomer matrix; domains and the matrix are linked by dense covalent bonds. b Upon deformation, the rigid nanodomains do not deform, and the dense covalent interlinks transmit stress. The architecture produces a crucial strain-amplification effect: macroscopic strains are greatly magnified in the soft matrix. Results Theoretical prediction: avoiding instability by strain amplification Consider a sheet of Ordered nanoPlastic–Elastomer Network (OPEN) under uniaxial deformation. Because the rigid nanodomains are orderly distributed in the matrix, we consider a repeat unit of two rigid nanodomains sandwiching a layer of elastomer matrix (Fig. 2a). The nanodomains act as rigid inclusions that deform negligibly, and covalent interlinks transmit stress. Consequently, the matrix experiences an amplified strain relative to the overall material. For a given volume fraction (φ) of the rigid domains, the undeformable fraction along one dimension scales as ~φ1/3. To a first approximation, the strain-amplification effect is expressed as15, where λ is macroscopic stretch applied to the material, and λm is microscopic stretch in the soft matrix. For an OPEN (φ > 0), strain is amplified, λm > λ, whereas for a conventional polymer network (φ = 0), Eq. (1) recovers the expression of affine deformation, λm = λ. Consequently, stiffening in the OPEN occurs at a smaller macroscopic stretch than in a conventional polymer network. Theoretical prediction. a A repeat unit in the ordered nanoplastic-elastomer network (OPEN) under uniaxial deformation. b A unit cell of OPEN with 5 vol.% of rigid plastic nanodomains. c Microscopic stretch as a function of macroscopic stretch during uniaxial tensile. d Finite element method (FEM) simulation result of an OPEN under uniaxial tensile. Stress-stretch curves are derived from the simulation under (e) uniaxial tensile and f uniaxial compression. g A DEA has a thickness H at the undeformed state. The electrostatic stress compresses the DEA to a thickness h. h The normalized electric field as a function of area expansion. i A balloon under inflation pressure. j The normalized pressure as a function of area expansion. Under uniaxial compression, the rigid nanodomains set a lower bound on compressive stretch along the thickness direction, beyond which the material stiffens sharply. This limiting compressive stretch scales as Consequently, the compression-stiffening in the OPEN also occurs earlier than in conventional polymer networks. We conduct a simulation through the finite element method (FEM) to visualize the strain amplification and predict the stress-stretch behaviors (Supplementary Information Text). The rigid spherical nanodomains are taken to be arranged in a single cubic packing under periodic boundary conditions. The soft matrix is described by the Arruda-Boyce model16, and is incompressible. The rigid nanodomains are taken to be much stiffer than the soft matrix. The rigid nanodomains and the soft matrix are assumed to be perfectly bonded. A unit cell is far smaller than the bulk material, such that the boundary effect is neglected. Consider a unit cell of an OPEN with φ = 0.05 (Fig. 2b). Apply a stretch to the unit cell, the local stretch at the midpoint between neighboring rigid nanodomains is taken as a representative microscopic stretch in the soft matrix. The relationship between the microscopic and macroscopic stretches approximately follows Eq. (1) (Fig. 2c). As shown in Fig. 2d, the strain is significantly amplified within the soft matrix. Nominal stress (s) scales with the shear modulus (μ) of the soft matrix, s ~ μ. Therefore, for OPENs with various φ, we derive the dimensionless stress s/μ as a function of macroscopic stretch, s(λ)/μ (Supplementary Fig. 1). The results reveal that incorporating 5 vol.% of rigid, well-bonded nanodomains induces a pronounced strain-amplification effect, preserving the softness at small strains while enhancing strain-stiffening at moderate to large strains under both uniaxial tension (Fig. 2e) and compression (Fig. 2f). By using the stress-stretch relationship derived from the finite element method, we further demonstrate that an OPEN can prevent snap-through instability in a dielectric elastomer actuator, and a pneumatic actuator. In a dielectric elastomer actuator (DEA), an elastomer membrane of thickness H is coated with two thin layers of flexible electrodes. The membrane has a dielectric constant of ε. A voltage U generates electrostatic stress that compresses the membrane to a thickness h (Fig. 2g). This electrostatic stress is refs. 3,17, The true stress of elastomer membrane is −λs(λ), where the negative sign exists because we define the tensile stress as positive. At each equilibrium state, force balance between true stress of the membrane and electrostatic stress gives −λsλ=εUHλ2, so that Denote the area expansion, λa, as the ratio of area of the elastomer membrane in the deformed state to that in the undeformed state, so that λa=1/λ. Because nominal stress (s) scales with the shear modulus (μ) of the soft matrix, s ~ μ, we can obtain the dimensionless electric field, UHεμ, as a function of area expansion, λa (Fig. 2h). For a conventional polymer network, φ = 0, the function is non-monotonic, and the DEA suffers from snap-through instability, during which the actuator breaks down. In fact, snap-through instability is one of the major failure modes of a dielectric elastomer actuator, and severely limits the actuation performance3,10,18–21. In contrast, for an OPEN with φ = 0.05, the function becomes monotonic, indicating that the DEA would avoid snap-through instability. This theoretical prediction will be tested experimentally in the following sections. Snap-through instability is also extensively found in pneumatic actuators, where a critical air pressure induces a sudden and discontinuous deformation of a soft elastomer membrane2. Consider a spherical balloon with a thin membrane of OPEN. In the undeformed state, the balloon has a radius R and a thickness H. When subjected to an internal pressure p, the balloon expands to a radius r, and the thickness decreases to ℎ (Fig. 2i). Denote the macroscopic thickness stretch as λ = h/H. In the deformed state, the equilibrium between the circumferential true stress, λs(λ), in the membrane and the internal pressure gives: −2πrhλsλ=πr2p, so that The elastomer membrane is taken to be incompressible, so that hr2=HR2, and Area expansion of the balloon is λa=1/λ. We obtain normalized pressure, pRμH, as a function of λa (Fig. 2j). For a conventional polymer network, φ = 0, the balloon undergoes snap-through instability, whereas for an OPEN, the material suppresses this instability, effectively stabilizing the balloon deformation. The design of OPEN requires the rigid domains to be much stiffer than the soft matrix, so that they deform negligibly and amplify strain. The modulus effect plateaus when the domain modulus is approximately three orders of magnitude greater than that of the matrix (Supplementary Fig. 2). The ability of the OPEN architecture to avoid instability does not rely on the specific tensile behavior of the soft matrix (Supplementary Fig. 3), which suggests a general design principle for soft materials that resist instability-induced failure. The effectiveness of the OPEN architecture is expected to be bounded by characteristic length scales spanning from the molecular to the macroscopic level. At the lower bound, the size of the rigid nanodomains should exceed the mesh size of the polymer network (typically a few to tens of nanometers), so that the domains can engage multiple network chains and induce effective strain amplification. At the upper bound, the domain size should remain much smaller than the structure's characteristic dimensions, e.g., the material's thickness or device dimensions, to preserve mechanical homogeneity and enable cooperative deformation. Within these bounds, the strain-amplification effect is governed primarily by the volume fraction of rigid domains and is, to first order, independent of their absolute size (Eqs. 1–2), while particle size influences secondary factors such as interfacial area and structural uniformity. Taken together, these considerations suggest a practical length scale for OPEN in the range of ~10 nm to 1 μm. Later, we will show a system that resides in this range. Preparation of ordered nanoplastic-elastomer network We introduce a method of preparing an ordered nanoplastic-elastomer network (OPEN) using colloidal assembly of a core-shell latex. The latex is prepared via living radical emulsion polymerization, mediated by a reversible addition-fragmentation chain transfer (RAFT) agent. The model system includes three species of monomers: a monomer of soft matrix, 2-ethylhexyl acrylate (EHA), a monomer of rigid plastic nanodomains, styrene (St), and a functional monomer with a photocrosslink group, 4-acryloyloxy benzophenone (ABP). We use a RAFT agent of an amphiphilic di-block oligomer, which also acts as a surfactant and forms micelles in water (Fig. 3a). EHA and ABP are added into the aqueous solution of the oligomeric RAFT agent, and polymerization is initiated by a thermal initiator at 70 °C. The copolymerization produces latex particles of uncrosslinked P(EHA-co-ABP) chains (Fig. 3b). The benzophenone groups act as latent photocrosslink groups. In a latex particle, each chain has one end of a diblock oligomer at the particle surface, serving as surfactant and stabilizing the latex, and has the other end of a RAFT group inside the particle, allowing for further chain growth. The molar ratio of ABP to EHA is designed to be 3 × 10-3, ensuring that each soft polymer chain has multiple photocrosslink groups. Next, styrene is added to the latex dispersion, initiating the growth of a polystyrene (PSt) segment at the RAFT end of each chain. Consequently, a core-shell particle form, with a glassy PSt core and a soft P(EHA-co-ABP) shell, covalently bonded via the RAFT polymerization (Fig. 3c). This core-shell structure is observed by a transmission electron microscope (Supplementary Fig. 4). Preparation of ordered nanoplastic-elastomer network (OPEN). a A RAFT agent of an amphiphilic di-block oligomer acts as a surfactant and forms micelles in water. b EHA and ABP are added into the aqueous dispersion of the RAFT agent, and polymerization is initiated by a thermal initiator. The reaction results in a soft latex of P(EHA-co-ABP). c St is added into the latex dispersion, yielding a core–shell latex consisting of a glassy PSt core and a soft P(EHA-co-ABP) shell. The core and shell are covalently linked. d The latex is dried and then exposed to ultraviolet light. The photocrosslinking groups crosslink the soft segments into a network. The resulting material is a membrane of OPEN. e Photo of the core-shell latex. f Photo of the membrane. The scale bars in Fig. 3e and f represent 10 mm. g Atomic force microscopy image of the cross-section. The scale bar is 500 nm. h SAXS pattern. Fig. 3e–h correspond to the material containing 15 vol.% of PSt. Pour the latex into a petri dish and dry at room temperature. As water evaporates, the core-shell latex comes into close contact. Particles have a narrow distribution of diameters ~ 165 nm (Supplementary Fig. 5), such that they are likely to form a structure of close packing of spheres. The capillary forces deform the soft shell, and the latex particles sinter into a coherent membrane. The uncrosslinked soft polymer segments diffuse across the particle boundaries to form a continuous matrix embedding rigid spheres of glassy polymer segments. The membrane is then stored in an oven of 60 °C for 6 hours to completely remove water (Supplementary Fig. 6). Subsequently, we expose the membrane to ultraviolet light for one hour. During UV exposure, the photocrosslink groups generate radicals (Supplementary Fig. 7), and OPEN is formed by crosslinking the soft polymer segments (Fig. 3d). We prepare OPENs with various volume fractions of rigid plastic nanodomains (φ). Take one with φ = 0.15. The core-shell latex is milky white (Fig. 3e), and the membrane of OPEN is translucent and light blue (Fig. 3f). The light blue color is a structural color originated by the light scattering of the ordered rigid nanodomains. The atomic force microscopy (AFM) image of the material indicates the ordered distribution of the nanodomains (Fig. 3g, Supplementary Fig. 8). The center-to-center distance between the neighboring cores, calculated from the first-order peak of the SAXS profile, is about 105 nm (Fig. 3h). A model of sphere form factor with a hard sphere structure factor fits well with the first-order peak of the SAXS curve, and suggests that the hard domains have a diameter ~60 nm (Supplementary Fig. 9), which is comparable to the observation of the microscopy. Although higher-order peaks are present in the SAXS profiles, which indicate an ordered structure, they are not sufficiently pronounced to indicate an ideally ordered structure of a quasi-crystal. We will show later that the mechanical behaviors of OPENs, however, match well with the FEM results based on the perfectly ideal order structure. This suggests that the mechanical behavior may be relatively insensitive to variations in interdomain spacing. AFM and SAXS profiles of OPENs with various φ are presented in Supplementary Figs. 8 and 9. Mechanical and electroactuation behaviors of OPEN Besides the OPENs with various volume fractions of rigid nanodomains, a conventional polymer network is synthesized by curing a precursor mixture of monomer, photoinitiator, and crosslinker (Supplementary Fig. 10). All the networks share the same soft polymer strands, PEHA, and have a comparable crosslink density. We stretch the four materials of φ = 0, 0.05, 0.10, and 0.15, and record nominal stress as a function of uniaxial stretch, s(λ) (Fig. 4a). As φ increases, the stiffening behavior occurs at smaller stretches. The experimental data dots coincide with the stress-stretch relation derived from the finite element simulation (material parameters are summarized in Supplementary Table 1). Interestingly, when we use Eq. (1), λm=1+λ−11−φ1/3, to replace the macroscopic stretches with microscopic stretches, all the curves overlap. It supports the idea that the strain-amplification effect indeed alters tensile behavior and advances strain stiffening in OPENs. Ordered nanoplastic-elastomer network avoids electromechanical snap-through instability. a Macroscopic stress–stretch curves under uniaxial tensile for OPENs with various volume fractions of rigid nanodomains, φ. b Microscopic stress-stretch curves by converting the curves in Fig. 4a by using Eq. 1. Area expansion as a function of nominal electric field for (c) regular polymer network, and d ordered nanoplastic-elastomer network with φ = 0.05. In Fig. 4c and d, the dots of each shape represent one separate experiment. For each network, three separate experiments are conducted. The scale bars in the photos are 1 mm. e Maximum area expansion, breakdown field, and maximum energy density of the two networks. Data with error bars in (e) are presented as mean values ± standard deviations (s.d.) of n = 3 replicates. f Maximum area expansion of OPEN as a function of volume fraction of rigid nanodomains. The modulus of the OPEN increases mildly as the volume fraction of the rigid domains increases (Supplementary Fig. 11). Young’s modulus is ~240 kPa for the soft matrix itself, and is ~280 kPa for the OPEN with φ = 0.05. That is, a small volume fraction of the rigid domains can enhance strain-stiffening while largely preserving the softness. Under cyclic loading, OPENs exhibit small hysteresis (Supplementary Fig. 12). The hysteresis further decreases over cycles (Supplementary Fig. 13), which is known as the shakedown behavior commonly observed in soft materials22. We stretch each material to twice its original length for 1000 cycles, the mechanical response is stable, and plateaus after ~100 cycles. The low hysteresis and cyclic stability can be attributed to three factors. First, the polymer chain in the soft matrix and well crosslinked into a network, such that most chains deform by entropic elasticity (Supplementary Fig. 14). Second, the noncovalent interactions between PEHA chains are weak, such that each cycle dissipates negligible energy. Third, the covalent interlinks between the matrix and the domains are strong, such that the soft matrix does not detach from the domains. We further demonstrate OPEN capable of preventing snap-through instability during dielectric actuation. We compare the actuation behaviors of an OPEN with φ = 0.05, and the conventional polymer network. Though the two networks only differ by 5 vol.% of rigid nanodomains in composition, the actuation behaviors differ dramatically. The conventional network shows a maximum area expansion of ~1.3, and breaks down at a nominal electric field of ~30 V μm⁻¹, corresponding to a true electric field of ~40 V μm⁻¹ (Fig. 4c). In comparison, the OPEN shows a maximum area expansion of ~2.1, and breaks down at a nominal electric field of ~45 V μm⁻¹, corresponding to a true electric field of ~100 V μm⁻¹ (Fig. 4d). The architecture of OPEN greatly enhances the actuation performance, especially, increasing the actuation energy density by one order of magnitude (Fig. 4e). The greatly enhanced actuation performance of OPEN can be well predicted by using the electromechanical coupling theory along with the stress-stretch relation derived from FEM, which are drawn as the solid curves in Fig. 4c, d. When drawing the theoretical curve, the material parameters used to describe the soft matrix are summarized in Supplementary Table 1, and the dielectric constant is measured by experiment (Supplementary Fig. 15, Supplementary Table 2). We also draw an electrical breakdown curve as the orange dash curve, by assuming the electrical breakdown strength of the material is ~90 V μm⁻¹. For the conventional network, it suffers from a failure mode of electromechanical snap-through instability: a critical voltage triggers a positive feedback loop between reduction of membrane thickness and increase of the true electric field. As a result, the actuator breaks down at an electric field much lower than the electrical strength of the polymer. In comparison, the OPEN avoids electromechanical snap-through instability, and the actuator achieves large, stable area expansion, breaking down at the electrical breakdown strength of the polymer. For OPENs with various φ, the maximum expansion during the actuation scales as λa, max ~ φ-0.30 (Fig. 4f, Supplementary Fig. 16). We understand this scaling relation as follows. When the electrical breakdown strength of the material is sufficiently high, the undeformable nanodomains sets the maximum area expansion. The limiting compressive stretch along the thickness direction scales as λlim~φ1/3 (Eq. 2), such that the maximum area expansion λa,max~φ−1/3. Counterintuitively, it indicates that if one desires a large area expansion during electroactuation, the volume fraction of the rigid nanodomains should be properly designed, rather than by continually increasing the volume fraction. The critical volume fraction of the rigid plastic nanodomains to suppress electromechanical snap-through could be less than 5 vol.%, depending on the mechanical behavior of the soft matrix, the packing of the domains, and the loading conditions (Supplementary Figs. 17–20). For instance, if the soft matrix itself stiffens at small stretches, the required volume fraction of rigid domains is reduced (Supplementary Fig. 17). However, promoting the strain-stiffening always requires to increase the crosslink density of the soft matrix, which leads to a network that is overly rigid and brittle, as we discussed in the introduction. Bottlebrush polymer networks provide an elegant solution to this trade-off between stability and softness23,24. The long side chains increase backbone rigidity of polymer chains to promote strain-stiffening, while simultaneously reducing the load-bearing chain density to maintain softness14,25,26. The bottlebrush elastomers have demonstrated their capability to suppress electromechanical instability23,24. In comparison, the architecture of OPEN follows a conceptually distinct yet complementary route. Rather than encoding strain-stiffening within the soft matrix's topology, OPEN introduces a small volume fraction of rigid domains that act as strain amplifiers, suppressing instability even when the matrix itself is weakly strain-stiffening, thereby allowing the mechanical response to be tuned in a modular manner. These two strategies are not mutually exclusive. It has been shown that incorporating rigid domains into bottlebrush networks, for example via block copolymer design, can further enhance strain-stiffening of bottlebrush networks23,24. This effect can likewise be partially interpreted as strain amplification by rigid domains. We emphasize here that the design of OPEN is fundamentally different from conventional elastomers reinforced by rigid particles. In the OPEN, rigid nanodomains are orderly distributed in the soft matrix, and are perfectly bonded to the matrix by covalent bonds. In contrast, in particle-reinforced elastomers, rigid particles are randomly distributed in the soft matrix, and are adhered with the matrix majorly by noncovalent interactions, sometimes by sparse covalent bonds. In a particle-reinforced elastomer, by randomly distributed, it means that the thickness of a soft matrix between adjacent rigid particles can be far smaller than the diameter of a particle. Consequently, upon stretching, the thin matrix between neighboring particles experiences triaxial stress and appear highly stiff27. At moderate deformations, however, the weak interfacial adhesion between the particles and the matrix often leads to debonding, causing the material to soften under strain28,29. Indeed, such particle-reinforced elastomers typically exhibit strain-softening rather than strain-stiffening, a phenomenon well-known as Payne effect30 and Mullins effect31. In the OPEN, by orderly distribution, it means the distances between neighboring nanodomains are comparable. When the volume fraction of the nanodomains is relatively low (<20 vol.%), the thickness of the soft matrix separating adjacent nanodomains is comparable to, and can be larger than, the diameter of a nanodomain, such that this soft matrix will not undergo triaxial stress and apparently remain soft. Being well bonded means that each rubbery chain in the soft matrix is linked to the nanodomain, such that stress is transmitted, and the rubbery chain will not detach from the nanodomains. We employ FEM simulation to highlight the importance of strong interfacial bonding. Two cases are simulated, one with perfect domain-matrix bonding, the other with weak domain-matrix bonding; both have the same volume fraction of the rigid domains (5 vol.%). Uniaxially stretch both unit cells to 3.5 times their original lengths, the strong interface let the rigid domains to act as undeformable inclusions that transmit stress (Fig. 5a), whereas the weak interface leads to debonding and cavitation (Fig. 5b). The strong interfaces lead to strain amplification, whereas the weak interfaces do not (Fig. 5c). The strong interfaces promote strain-stiffening, whereas weak interfaces cannot (Fig. 5d). Validation of the design principles behind OPEN. In FEM simulation, stretch (a) a material of strong interface bonding, and (b) a material of weak interface bonding, to 3.5 times its original length. c Microscopic stretch as a function of macroscopic stretch. The local stretch at the midpoint between neighboring rigid nanodomains is taken as a representative microscopic stretch. d Stress-stretch curves derived from the FEM simulation. e Schematic of the design of random nanoplastic-elastomer network (RPEN) and ordered nanoplastic-elastomer network (OPEN). f Load both materials to twice original length and unload. g The strong bonding in OPEN prevents rubbery chain to detach from the rigid domains, while the weak bonding in RPEN cannot prevent detachment. h Performance of RPEN versus OPEN. Data with error bars in (h) are presented as mean values ± standard deviations (s.d.) of n = 3 replicates. Experimentally, we highlight the principle of OPEN by designing a control network, called random nanoplastic-elastomer network (RPEN), in which rigid plastic nanodomains are randomly embedded within the soft elastomer matrix, and the rigid domains adhere to the soft matrix only by weak noncovalent interactions (Fig. 5e). To prepare RPEN, we separately prepare a latex of P(EHA-co-ABP) and a latex of PSt and then mix the two latexes. The diameter of the PSt latex is designed to be ~ 60 nm, which is comparable to the domain size in OPEN (Supplementary Fig. 21). The mixed latex is dried into a membrane. Upon drying, the particles of dissimilar polymers form close contact and pack randomly. The P(EHA-co-ABP) particles flow into a continuous soft matrix, while the PSt nanodomains maintain their shapes. Upon UV exposure, the soft matrix is crosslinked into a network by covalent bonds, while the PS nanoparticles adhere to the soft matrix by noncovalent interactions. The random distribution of the PS nanoparticles is confirmed by AFM (Supplementary Fig. 22). To ensure a meaningful comparison, the RPEN is formulated to match the OPEN in both crosslink density and nanodomain fraction. Compared to OPEN, RPEN shows distinct stress-stretch behaviors. The RPEN shows a higher modulus (~550 kPa) than the OPEN ( ~280 kPa). In the OPEN, the nanoplastic domains are orderly embedded in a soft matrix, with rigid domains separated by it, which maintains a low modulus. In the RPEN, the nanoplastic domains are randomly packed, such that rigid domains may form close contact, which greatly increases modulus32–34. Load both networks to a stretch of two and unload, the RPEN exhibits a pronounced hysteresis (~22%), whereas the OPEN shows negligible hysteresis (Fig. 5f). Upon a moderate stretch, in the OPEN, the dense and covalent particle-matrix interlinks prevent detachment and maintain elasticity. However, in the RPEN, the weak noncovalent particle-matrix interactions lead to detachment and contribute to hysteresis (Fig. 5g)35. Under external voltages, the RPEN has a substantially reduced area expansion (λa ~ 1.4), only about half that of the OPEN (λa ~ 2.1) (Supplementary Fig. 23, Fig. 5h). Moreover, we emphasize that OPEN architecture is a modular design which enhances strain stiffening, and maintains elasticity and softness, while the specific choice of the soft matrix would introduce molecular-dependent characteristics. To demonstrate this, we prepare an OPEN system using poly(ethyl acrylate) (PEA) as the soft matrix with 5 vol.% polystyrene (PSt) as rigid domains (Supplementary Fig. 24). PEA has a higher entanglement density and stronger intermolecular interactions than PEHA, yet these interactions remain sufficiently weak at room temperature to enable low hysteresis. The pure PEA elastomer exhibits a modulus of ~780 kPa, a low hysteresis of ~5%, and weak strain stiffening. In contrast, the PEA-based OPEN maintains the modulus of ~830 kPa, the low hysteresis of ~10%, and significantly enhanced strain stiffening. Discussion By both theoretical modeling and experiment, we demonstrate a polymer network design—called the ordered nanoplastic–elastomer network (OPEN)—that avoids snap-through instability by amplifying microscopic strain. The design requires embedding ~5 vol.% of rigid plastic nanodomains into a soft elastomer matrix and covalently linking the domains and the matrix. Experimentally, we show that this network prevents premature electrical breakdown caused by electromechanical snap-through instability in a dielectric elastomer actuator and greatly enhances its performance by one order of magnitude. We also show the potential of this network in avoiding snap-through instability of pneumatic actuators. While embedding rigid domains with strong interfacial bonding is an old wisdom in composite design, it has primarily been used to enhance stiffness, strength, and toughness. In contrast, this work aims to maintain softness at small strains while promote strain-stiffening at moderate strains. This unusual combination of properties enables a distinct mechanical function—suppression of snap-through instability, which is not a primary objective in conventional composite design, but is important for many functional devices. We acknowledge that strain-amplification effect (Eq. 1) can be traced back to classical composite theories in the 1960s28, however, to our best of knowledge, it is the first time to design a material system to make use of this effect in benefitting functional soft devices. Indeed, many conventional nanocomposites including, filled rubbers and thermoplastic elastomers, exhibit increased stiffness but often show strain-softening rather than strain-stiffening at moderate strains29,36,37. The unusual mechanical response of the OPENs arises from a set of design principles, which include the incorporation of rigid domains that deform negligibly relative to a soft matrix, a controlled spatial distribution that enables cooperative deformation, strong covalent interfacial bonding to ensure efficient stress transfer, and a small volume fraction of rigid domains to balance strain-stiffening with overall softness. These principles are not restricted to the specific chemistry used here and are expected to be broadly applicable across various material systems. In this work, we use the colloidal assembly of a core-shell latex system to demonstrate the concept. Similar behaviors may be achieved through alternative approaches such as the self-assembly of block copolymers38, microphase separation of polymer hybrids39, coupling chemistries40,41, microgel embedding42,43, and lithography44, etc. At the same time, the specific material choices introduce system-dependent characteristics that are important to specific applications. For example, the glass transition temperatures of the rigid and soft phases can determine the stiffness contrast, the dielectric constant affects electromechanical performance, and the topology and noncovalent interaction of the soft matrix influence hysteresis. These factors govern the quantitative behavior of the material, but should not alter the underlying mechanism of strain amplification enabled by the architecture design. The mechanism identified here also provides a framework to understand recent experimental observations in related systems. For example, enhanced strain-stiffening has been reported in nanostructured hydrogels and elastomers containing rigid domains covalently coupled to soft matrices25,40. These behaviors can be partially understood in terms of strain amplification by rigid domains. More broadly, the strain-amplification effect is also recently noticed to be a crucial factor in contributing to the strain-stiffening behavior of soft tissues45–47. We believe OPEN provides a fertile model system for studying this kind of strain-amplification effect and shedding light on biophysics. A key advance of OPEN lies in its ability to combine softness with rapid compression-stiffening—two properties traditionally considered mutually contradictory in polymer networks. This unusual combination extends the implications of OPEN beyond soft actuators. For example, a soft seal should be soft enough to fill gaps between hard materials and roughness on surfaces, while strain-stiffening is required to bear fluid pressure48. For another example, many biological tissues, such as skins, are soft at small deformations to enable flexibility, and pronouncedly strain-stiffening at larger deformation to prevent injury25,26. OPEN may inspire engineering synthetic materials with skin-like nonlinear elasticity. Further opportunities include introducing anisotropy by aligning domains into non-spherical geometries, which may enhance directional strain amplification and strain-stiffening behavior (Supplementary Fig. 25). Method Materials 2-Ethylhexyl acrylate (EHA, > 99%, Aladdin) and styrene (St, > 99%, Aladdin) is purified by vacuum distillation to remove the inhibitor. 4-benzoylphenyl acrylate (ABP, > 98%, Aladdin), ammonium persulfate (APS, > 99%, Macklin), ammonium hydroxide (NH3·H2O, AR), 2-Hydroxy-4’-(2-hydroxyethoxy)-2-methylpropiophenone (2959, > 99%, Sigma-Aldrich) and 1, 4-butanediol diacrylate (BDDA, > 90%, Aladdin) are used without further purification. The macro-RAFT agent (polyacrylic acid-b-polybutyl acrylate prepared by RAFT polymerization) (PAA20-PBA5-RAFT) is provided by Entron. Single-wall carbon nanotubes (SWCNTs) for electrode fabrication are purchased from Aladdin Ltd. Preparation of the ordered nanoplastic-elastomer network (OPEN) We first prepare a seed latex of P(EHA-co-ABP). Firstly, 2.43 g of macro-RAFT, 29.88 g of EHA, 0.12 g of 4-ABP, and 129.79 g of deionized water are mixed in a 250 mL flask. The reaction system is purged with nitrogen for 30 min at room temperature to remove oxygen. The flask is then placed in a water bath at 70 °C, with stirring set at 250 rpm. An APS aqueous solution (0.015 g APS in 0.29 g deionized water) is injected as the initiator to start the reaction. The reaction lasts for 180 min. Afterwards, an NH3·H2O aqueous solution (1.85 g NH3·H2O in 6.58 g deionized water) is added slowly at 1 g min−1 to stabilize the latex. We take the ordered nanoplastic-elastomer network as an example to illustrate the subsequent material preparation. 38.32 g of the seed latex, 35.91 g of EHA, 0.15 g of ABP, and 74.88 g of deionized water are mixed in a 250 mL flask. The mixture is purged with nitrogen for 30 min at room temperature and then placed in a 70 °C water bath. An APS aqueous solution (0.01 g APS in 0.20 g deionized water) is injected to initiate the polymerization. The reaction lasts 180 min. Subsequently, 2.48 g of St is added to the flask, and the polymerization of PSt segments lasts for 180 min. The latex is cooled down to room temperature and stored in a fridge at 5 °C. 2 g of latex is mixed with 3 g of deionized water, and is poured into a plastic petri dish with a diameter of 5.6 mm. The petri dish is covered with a layer of filter paper, and is placed on a hot plate at 35 °C for 24 h to evaporate water and form a membrane. The membrane is then stored in an oven at 60 °C for 6 hours to further remove the residual water. The fully-dried membrane is transferred into a UV chamber (UVGO, 3535 KXD, 200 W, 365 nm), and exposed to UV for 1 hour. Preparation of the conventional polymer network First, 10.00 g EHA, 0.0024 g 2959, and 20 μl BDDA are mixed and sonicated for 30 min (KQ-100E). The resulting solution is poured into a reaction cell composed of two glass plates separated by a silicone rubber spacer (~550 μm thick), and the two plates are clamped with binder clips. The assembled cell is placed inside a polyethylene press-to-close bag and exposed to UV light in a UV chamber for ~4 hours. The resulting elastomer membrane is stored in a fume hood overnight to evaporate the residual monomers. Preparation of random nanoplastic-elastomer network (RPEN) We separately prepared a latex of P(EHA-co-ABP), and a latex of PSt. For the latex of P(EHA-co-ABP), 40.63 g of seed latex of P(EHA-co-ABP), 37.75 g of EHA, 0.15 g of 4-ABP, and 73.45 g of deionized water are mixed in a 250 mL flask. The mixture is purged with nitrogen for 30 min at room temperature and then placed in a 70 °C water bath. An APS aqueous solution (0.01 g APS in 0.20 g deionized water) is injected to initiate the polymerization. The reaction lasts 180 min. For the PSt latex, 10.94 g of macro-RAFT, 45.02 g of St, and 130.78 g of deionized water are mixed in a 250 mL flask. The mixture is purged with nitrogen for 30 min at room temperature to remove oxygen. The flask is then placed in a water bath at 70 °C, with stirring set at 250 rpm. An APS aqueous solution (0.07 g APS in 1.30 g deionized water) is injected as the initiator to start the reaction. The reaction lasts for 180 min. Afterwards, an NH3·H2O aqueous solution (8.32 g NH3·H2O in 29.16 g deionized water) is added slowly at 1 g min-1 to stabilize the latex. The latex of P(EHA-co-ABP) and the latex of PSt are mixed at a volume ratio of 19:1. 2 g of mixed latex is mixed with 3 g of deionized water, and is poured into a plastic petri dish with a diameter of 5.6 mm. The petri dish is covered with a layer of filter paper, and is placed on a hot plate at 35 °C for 24 h to evaporate water and form a membrane. The membrane is then stored in an oven at 60 °C for 6 h to further remove the residual water. The fully-dried membrane is transferred into a UV chamber (UVGO, 3535 KXD, 200 W, 365 nm), and exposed to UV for 1 h. Preparation of SWCNT electrodes 10 mg SWCNT and 1.8 g SDS are mixed in 180 g deionized water and then ultrasonicated (300 W of power) for 10 min. SWCNT dispersions are obtained from the supernatant after centrifugation and decantation. The dispersion solution is then suction filtered through PVDF microporous membrane with a 0.45 μm pore diameter to form a SWCNT layer on the membrane ready for the transfer printing later. The sheet resistance of the electrodes is around 1.5 kΩ/sq measured by a four-probe setup. Tensile test Tensile tests are conducted using a universal testing machine (UTM 2502, SUNS Technology Co., Ltd., China). Elastomer sheets are cut into dumbbell-shaped specimens (width: 2 mm; gauge length: 12 mm) using a standard cutting device. Each specimen is clamped at both ends, and two horizontal reference lines are marked on the surface with a black marker. The samples are stretched at a constant strain rate of 0.05 s-1 while being simultaneously recorded using a camera. The distance between the reference lines at different times is analyzed with ImageJ to obtain the time–strain relationship and corresponding stress–stretch curve. Modulus is defined as the initial slope of a stress–stretch curve (stretch < 1.05). Electro-actuation test Electro-actuation performance is conducted using a diaphragm actuator at a prestrain-free condition according to an accepted standard49. The elastomer films are placed into two rigid acrylic rings with 10 mm inner diameter, and the bottom ring stuck tightly onto the top of a diaphragm chamber. A small bias air pressure of ~500 Pa is applied to guide the out-of-plane actuation and to prevent wrinkling. The diameter of the active area is 6 mm (before the actuation and the set of bias air pressure) with two SWCNT electrodes on both sides of an elastomer film. A high-voltage power supply (Trek 610E) is used to drive the actuators, and the voltage (U) is steadily increased until dielectric breakdown occurs. Images and video of actuation are obtained by a camera (Canon 70D). The video and images are analyzed using Adobe Premiere Pro and Adobe Photoshop, and the area strain (λa) generated by dielectric actuation is calculated according to λa=r2+h2r02+h02 (Supplementary Fig. S26) The maximum energy density is calculated according to the following formula: emax=12ε0εEb2lnλa,max, where ε0 is the vacuum permittivity (8.85×10-12 F m-1), and ε is the relative dielectric constant of the elastomer measured at 1 kHz. The true electric breakdown field, Eb, and the maximum area expansion, λa,max, are obtained from the actuation tests. Transmission electron microscopy (TEM) The OPEN latex is diluted 50 times with deionized water, and a drop of the diluted dispersion is deposited onto a carbon-coated copper grid. After drying at room temperature, the sample is stained with RuO4 vapor for 2 hours to enhance contrast, and then the morphology of the latex particles is observed using a transmission electron microscope (TEM, HT-7700, Hitachi, Japan). Atomic force microscopy (AFM) Atomic force microscopy of the cross-section of each material is performed on a Bruker Dimension Icon XR system (Bruker Corporation, USA) in a tapping mode, using an RFESP-75 probe (Bruker Corporation, USA). The sample is cut at -120 °C using a Leica EM UC7 ultramicrotome (Leica Microsystems, Germany), to expose a smooth cross-section. Atomic force microscopy of the surface of each material is performed on a Cypher S AFM Microscope (Oxford Instrument) in an AC Air Topography, using an 160AC-NA probe (OPUS by MikroMasch). The sample is prepared by casting latex on a silicone water. Dynamic light scattering (DLS) The hydrodynamic diameter of the latex particles is measured using a dynamic light scattering (DLS) analyzer (Zetasizer Nano-ZS, Malvern Instruments, UK) at 25 °C. The latex dispersion is diluted with deionized water to a concentration of 1 mg mL-1 prior to measurement to minimize multiple scattering effects. Each sample is measured three times, and the average value of the hydrodynamic diameter is reported. Thermogravimetric analysis (TGA) Thermogravimetric analysis (TGA) is performed using a thermogravimetric analyzer (Q500, TA Instruments, USA). Approximately 8-10 mg of dried elastomer samples are heated from 50 °C to 600 °C at a rate of 10 °C min-1 under a nitrogen atmosphere. The mass change is recorded as a function of temperature. Dielectric constant measurement The dielectric constant of the elastomer film is measured using a broadband dielectric spectrometer (Concept 40, Novocontrol Technologies, Germany) at 25 °C over a frequency range from 10-1 Hz to 107 Hz. The films are cut into circular discs with a diameter of 20 mm. Each disc is sandwiched between two 20 mm copper electrodes under slight pressure to ensure good contact. Supplementary information Supplementary Information Transparent Peer Review file Source data Source Data Source Data Acknowledgements The authors would like to thank Jing He for her assistance with atomic force microscopy, and Sudan Shen for her assistance with ultrathin frozen sectioning and transmission electron microscopy at the State Key Laboratory of Chemical Engineering and Low-Carbon Technology, Zhejiang University. Author contributions Z.C. conceived the idea and designed the study. Y.W., Z.L., and Z.C. conducted the experiments. Z.C. and G.N. did the theoretical analysis. G.N. conducted the finite element method simulation. All the authors contributed to the discussion. Z.C., Y.W., G.N., and Y.L. wrote the manuscript. Z.C., Y.L., G.N., and S.Q. supervised the project. Peer review Peer review information Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available. Funding Y. L. would like to acknowledge the financial support from the National Natural Science Foundation of China (Grant No. 21636008), the Science & Technology Program of Zhejiang Province (Grant Nos. 2021C01184, 2022C01034), and the Double Thousand Plan of Jiangxi Province. G. N. would like to acknowledge the financial support from the National Natural Science Foundation of China (Grant No. 12502199). S. Q. would like to acknowledge the financial support from the National Natural Science Foundation of China (Grant No. 12321002). Data availability The data in this study are available in the main article, supplementary information, and source data file. All data underlying this study are available from the corresponding author upon request. Source data are provided with this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Zheqi Chen, Email: zqchen@zju.edu.cn. Guodong Nian, Email: gnian@zju.edu.cn. Yingwu Luo, Email: yingwu.luo@zju.edu.cn. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-73807-5. References Associated Data Supplementary Materials Supplementary Information Transparent Peer Review file Source Data Source Data Data Availability Statement The data in this study are available in the main article, supplementary information, and source data file. All data underlying this study are available from the corresponding author upon request. Source data are provided with this paper.