2D MXene ‐Enabled Unidirectional Flexible Optoelectronic Memristor Toward Energy‐Efficient Neuromorphic Photosensing

Integrated optoelectronic synapses that combine sensing, memory, and computation in a mechanically compliant platform are attractive for energy‐efficient wearable artificial vision and in‐sensor computing. However, simultaneously achieving intrinsic rectification, analog weight programmability, optical responsivity, and facile, scalable fabrication remains challenging. Here, we report a self‐rectifying, flexible, semi‐transparent optoelectronic memristor fabricated by layer‐by‐layer solution processing and spray coating, thereby avoiding the use of vacuum‐deposited noble‐metal top electrodes. The indium tin oxide/ZnO/TiO ₂ nanosheet/MXene heterostructure uses defect‐rich ZnO as a photoactive oxygen‐vacancy reservoir, crystalline TiO ₂ nanosheets (TS) as a charge‐trapping and barrier‐modulating interlayer, and a termination‐rich 2D MXene sheets (MS) electrode to promote asymmetric carrier injection. The device exhibits unidirectional analog switching with a rectification ratio of ~10 ³ , stable potentiation and depression with nonlinearity values of 2.02 and 2.38, a conductance on/off ratio of 121.4, and 94.3% inference accuracy using experimentally extracted weight states. Under illumination, it shows wavelength‐selective photosynaptic behavior, stronger excitation at 365 nm, along with intensity‐ and duration‐dependent transitions from short‐term plasticity (STP) to long‐term plasticity (LTP), paired‐pulse facilitation (PPF), and learning‐forgetting‐relearning characteristics. These electrical and optoelectronic functions are preserved under bending, supporting robust mechanical reliability. Mechanistically, the response is governed by interface‐controlled barrier modulation mediated by oxygen‐vacancy redistribution, charge trapping/de‐trapping at the ZnO/TS interface, and asymmetric carrier injection at the MS/TS contact, as evidenced by postsynaptic current (PSC). This work establishes a scalable, semi‐transparent platform for energy‐efficient neuromorphic photosensing and in‐sensor wearable computing for artificial vision systems.