Jinsong Leng

Transparent polymer substrates for flexible perovskite solar cells and foldable displays must offer high transparency, thermal stability, and active deformability. With the rapid development of intelligent and adaptive flexible electronics, shape memory functionality has emerged as a key requirement for next-generation devices to enable smart behaviors such as self-deployment or morphing. However, although conventional colorless transparent polyimides exhibit outstanding transparency and thermal resistance, they tend to lack shape memory functionality, making it difficult to meet the demands of next-generation intelligent flexible devices such as self-deployable structures. We designed a molecular system using two dianhydrides, one containing a hexafluoroisopropylidene group and one containing ether linkages, and three fluorinated diamines with different rigidities. The hexafluoroisopropylidene and ether units suppress charge-transfer complex formation to enhance optical transparency, while the ether linkages also improve chain flexibility. The varying diamine rigidity enables tuning of thermal properties. Based on this design, colorless shape memory polyimide (co-CSMPI) films were prepared via quaternary copolymerization. The results demonstrate that co-CSMPIs exhibit excellent overall performance. A representative co-CSMPI possesses a glass transition temperature ( T _g ) of 241 °C, satisfying the requirements for high-temperature processing; a transmittance of 88.0% at 500 nm, indicating good optical transparency. Furthermore, it exhibits excellent shape memory performance, with shape fixity ratios ( R _f ) of 99.4% and shape recovery ratios ( R _r ) of 98.9%. The colorless shape memory polyimide films enable lighter device designs, holding great promise for flexible perovskite solar cell substrates and foldable display covers.

While self‐powered flexible e‐skins based on nanogenerators hold immense promise for wearable electronics, severe interfacial mechanical mismatches occur under complex joint deformations, limiting reliable multichannel sensing. Herein, a 4D‐printed mechanical metamaterial‐enabled e‐skin with an auxetic scaffold and AI‐assisted decoding is developed to improve deformation accommodation and support multichannel sensing during joint motion. Benefiting from an auxetic architecture with a tunable negative Poisson's ratio (−0.41 to −1.76), the metamaterial skeleton expands laterally under stretch to mimic the strain‐stiffening behavior of biological tissues. By concentrating local stress, this geometrical programmability transforms the skeleton into an active mechanical amplifier, significantly boosting the triboelectric‐piezoelectric (TENG/PENG) responses. An integrated nine‐channel array captures these amplified spatiotemporal features with high fidelity. During fast running, the device yields robust 3 V outputs and resolves 3 Hz rhythmic motions. Further, by introducing a lightweight 1D convolutional neural network (1D‐CNN) to decouple intricate spatiotemporal crosstalk, the system achieves advanced motion‐state classification with an exceptional accuracy of over 95% on average across different tasks. This integration of mechanics and artificial intelligence offers a robust platform for next‐generation intelligent wearable rehabilitation and human‐machine interfaces.

This paper presents a dual-mode pneumatic artificial muscle integrated with three circumferentially arranged S-shaped shape memory polymer (SMP) constraints. By selectively heating the SMPs, the actuator achieves reversible switching between axial elongation and multi-directional bending, enabling four distinct deformation modes (3C, 2C1H, 1C2H, and 3H). A thermo-mechanically coupled, energy-based theoretical framework is developed to model this behavior. The static equilibrium equations are derived from the total potential energy functional—which includes the hyperelastic strain energy of the bladder, the temperature-dependent stiffness of the SMPs, and the work done by pneumatic pressure—using the principle of minimum potential energy. A high-fidelity finite element model employing a sequentially coupled thermo-pneumatic-solid strategy is also established. A prototype was fabricated and systematically tested on a custom platform integrating pneumatic, thermal, and mechanical measurement systems. The results demonstrate good agreement among the theoretical model, finite element simulation, and experimental data, with the majority of relative errors falling below 10%. The experimentally measured maximum bending curvature reaches 0.0174 mm q − 1 , and the maximum axial elongation ratio is about 18.4%. An error analysis identifies residual SMP stiffness, contact friction, and braided sleeve slippage as primary sources of discrepancy. This study provides a systematic theoretical and technical basis for designing, modeling, and controlling multi-modal soft actuators.

The increasingly complex tasks of intelligent equipment have shifted the demand for structural materials from mere lightweight protection to multifunction integration, i.e., high recoverable energy absorption and accurate monitoring. Herein, we presented a material-architecture-manufacture integrated strategy for piezoelectric metamaterials to simultaneously achieve mechanical protection and self-powered sensing. Specifically, an intercalation heterostructure of silane-modified barium titanate (K-BTO)/MXene was constructed to improve the interfacial polarization effect, and an innovative electric field–assisted 4D printing technology could effectively induce the in situ poling of K-BTO. Furthermore, the Kagome lattice–inspired metamaterial (k-CAH) exhibited superior specific energy absorption (0.18 joules per gram), resulting from compression-bend-torsion coordination and multidirection coupling mechanisms. The multimode coupling deformation and local strain amplification mechanism, induced by the geometry design, substantially improved output capacities through activating various piezoelectric response modes at low frequency. With the material-architecture-manufacture synergy, g 33 eff reached up to 1.712 volt-meters per newton. The developed piezoelectric metamaterial system exhibited accurate self-powered sensing of real-time impact and high-efficiency energy harvesting under microvibration conditions, promising for the next-generation smart structural materials.

The mechanical response of history-dependent materials depends strongly on loading paths and thermomechanical histories, making neural network constitutive modeling inherently challenging. Most existing approaches emphasize fitting complete historical sequences, resulting in excessive parameters, limited generalization, and an inability to initiate predictions from arbitrary intermediate states or variable time steps. This study theoretically demonstrates that for compressible history-dependent materials (CHDMs), the evolution of internal variables exhibits Markovian behavior, revealing structural redundancy in conventional sequence-based neural networks. Accordingly, a Local Decoupling Evolution Operator Network (LDEONet) is proposed, which explicitly decouples internal variable evolution into direction and intensity components. Shape memory polymers (SMPs) are employed as a representative case study. Experimental results show that LDEONet requires only approximately 101 parameters and achieves high accuracy using just 1/175 of the training samples required by long short-term memory networks (LSTM). Crucially, LDEONet enables both path and discretization generalization, allowing predictions from arbitrary intermediate states while maintaining numerical stability under constant, random, sinusoidal, and stepwise discretization schemes. Failure boundaries under cross-modal thermomechanical transitions are also examined. LDEONet thus provides a lightweight, high-generalization paradigm for history-dependent material modeling based on local evolution principles.

High‐performance conductive gels possess exceptional mechanical robustness yet exhibit compromised ductility, conductivity, and self‐reinforcement capability, owing to the impeded ion transport and restricted dynamic reconstruction within dense networks. This dilemma is overcome here by a stress‐triggered ionic rearrangement strategy that exploits the dynamic reorganization of mobile ions to decouple ionic conductivity from structural rigidity. This strategy engineers a network saturated with dispersed free ions to guarantee superior conductivity independent of structural density, triggers in situ aggregation of these ions into high‐density ionic cross‐linking domains upon deformation, and consequently activates a potent self‐strengthening mechanism that reaches a maximum of 185% of the initial value. Consequently, this structural evolution translates into an unprecedented convergence of mechanical robustness and functional agility, yielding a tensile strength of 34.58 MPa, an elastic modulus of 77.38 MPa, and a volumetric toughness of 177.3 MJ/m ³ . Crucially, such extreme mechanical reinforcement is achieved without compromising the dynamic nature, preserving a high extensibility of 644.5% alongside efficient ionic conductivity. Leveraging this unique material platform, we develop a flexible sensor with a detection limit of 10 µm. The sensor demonstrates precise recognition of diverse physiological activities, from large‐scale limb movements to subtle signals such as swallowing, pulse, and respiration.