Neuromorphic sensing and computing in a versatile thermally grown Fe-W-O-S nanocomposite memristor

Memristive devices that combine rich plasticity with low power consumption are critical for neuromorphic hardware, yet synaptic, nociceptive, memory, and computing functions are often realized in separate elements. Here, we report a Fe-W-O-S nanocomposite memristor thermally derived from a solution-processed precursor that integrates these neuromorphic functionalities within a single cell. Structural and spectroscopic analyses using XRD, HRTEM with FFT, STEM-EDS, and XPS reveal a mixed Fe-W-O-S framework interfaced with α-Fe₂O₃. Oxygen and sulfur vacancies, together with mobile Ag⁺ ions, form coupled ionic and electronic conduction pathways that enable both diffusive and filamentary switching dynamics. The device operates at low voltage with typical DC SET and RESET voltages below 0.5 V and exhibits coexisting threshold switching and stable nonvolatile resistive switching with an ON/OFF ratio of approximately 10² and an energy consumption of about 50 pJ per switching event. Tailored identical and incremental pulse schemes enable analog long-term potentiation and depression with nonlinearity factors approaching unity and a coefficient of variation as low as approximately 2.7%. Paired-pulse facilitation, spike-number-dependent plasticity, and controllable transitions between short-term and long-term memory are demonstrated, together with nociceptive behaviors including threshold sensing, sensitization, hyperalgesia, and allodynia. Pavlovian associative learning and in-memory AND and OR logic operations are realized using the same device. By exploiting sixteen well-separated conductance states and intrinsic volatile dynamics, the Fe-W-O-S memristor operates as both a physical reservoir and a programmable weight element. When interfaced with a convolutional neural network readout, it achieves classification accuracies of approximately 97.9% for MNIST and 88.1% for Fashion-MNIST. These results establish Fe-W-O-S nanocomposite memristors as compact and energy-efficient building blocks for multifunctional neuromorphic sensing and computing systems.