Stretchable memristors are anticipated to be pivotal in the development of next-generation wearable and implantable electronic devices. However, the challenging fabrication requirements and restricted functionality have impeded progress in this field. In this Letter, ultraflexible and stretchable PDMS/Au/Ti-HQ/Al-HQ/Ag (denoted as ATP) memristor is fabricated using a pre-stretch–release process and molecular layer deposition. The ATP memristor exhibits stable bipolar resistive switching behavior at a curvature radius of 0.1 mm or under a 15% strain. The device features low switching voltages (VSET = 0.93 V and VRESET = −0.65 V), long-term retention performance (up to 104 s), stable endurance (up to 3000 cycles), and excellent multilevel conductance states. As an artificial synapse, it can simulate various important synaptic behaviors, such as long-term potentiation/depression, paired-pulse facilitation/depression, and spike time-dependent plasticity. Furthermore, basic logic and arithmetic operations have been realized using the ATP memristor, and parity checkers and encryption matrix systems have been constructed to ensure the accuracy and security of data transmission. Additionally, the ATP memristor-based deep convolutional neural network provides high accuracy and speed in image recognition tasks, achieving an 92% accuracy rate after training for 30 epochs. This cross-disciplinary research into memristor technology covers flexible electronics, digital circuits, information security, and brain-inspired computing, laying the groundwork for the next generation of high-performance, low-power artificial intelligence systems.
Ma et al. (Mon,) studied this question.