DIGITAL.CSIC. Repositorio Institucional del CSIC
Publicación
Artículo científico (JournalArticle)
Fast prototyping of memristors for ReRAMs and neuromorphic computing
The growing demand for energy-efficient computing in artificial intelligence requires novel memory technologies capable of storing and processing information. Memristors stand out in thanks to their ability to store information, mimic synaptic behavior and support in-memory computing architectures while requiring minimal active areas and energy consumptions. Here is presented a scalable and cost-effective approach to fabricate Ag/MoS2/Au memristors as resistive switching memory devices by combining roll-to-roll mechanical exfoliation of two-dimensional materials with inkjet printing. These devices exhibit reliable non-volatile switching behavior attributed to the formation and dissolution of metallic conductive filaments within the MoS2 layer, with high resistance ratios and robust retention times. A fully-connected neural networks is simulated using quantized weights mapped onto a virtual memristor crossbar array demonstrating that classification tasks can be performed with high accuracy even with limited bit-width precision, highlighting the potential of these devices for energy-efficient, high-throughput AI hardware. This journal is © The Royal Society of Chemistry, 2026, Authors gratefully acknowledge financial support from the ERC SKIN2DTRONICS (Contract Number 101167218) and the ENERGIZE (Contract Number 101194458) projects. Authors gratefully acknowledge also the Italian Ministry of Education and Research (MIUR) in the framework of the FoReLab project (Departments of Excellence) and the Piano Nazionale di Ricerca e Resilienza - PNRR (CN00000013). This work was also funded by the Ministry of Science and Innovation (Spain) through the projects (PDC2023-145920-I00 and PID2023-151946OB-I00) and the Severo Ochoa Centres of Excellence Program through Grant CEX2024-001445-S and the European Research Council through the ERC-2024-PoC StEnSo (grant agreement 101185235). The graphical abstract and Fig. 1, 2, and 4 were created, in part, with Blender via Blender Foundation., Peer reviewed