Molecular 2D/Polymer-based Memristors

Descripción del proyecto

Los memristores híbridos adaptables en tiempo real imitan las respuestas sinápticas a corto plazo

Un memristor (resistencia de memoria) funciona como una resistencia con memoria incorporada. Cambia su resistencia para regular la corriente basándose en su «memoria» de la tensión que se le ha aplicado anteriormente, y esa memoria se mantiene incluso cuando se desconecta la alimentación. Los memristores convencionales basados en óxidos inorgánicos están diseñados para mantener su estado a largo plazo. El proyecto Mol-2DMem, financiado por el Consejo Europeo de Investigación, pretende crear memristores híbridos que superen a los memristores basados en óxidos —capaces de funcionar de forma adaptativa, imitando las respuestas sinápticas a corto plazo— mediante la incorporación de especies iónicas en polímeros moleculares que contienen materiales bidimensionales. Los memristores allanan el camino al procesamiento de señales en tiempo real similar al humano para la visión o el reconocimiento del habla. Mol-2DMem integrará los memristores híbridos en módulos de preprocesamiento programables y presentará pruebas de concepto.

Objetivo

Our previous ERC project showed that 2D materials, when combined with molecular components, can be chemically tuned to unlock entirely new physical properties. This molecular/2D synergy enabled unprecedented control over electronic states and interactions, opening the door to transformative device concepts. Building on these breakthroughs, we now propose a new generation of hybrid memristors. By embedding ionic species within molecular polymers containing 2D materials such as MoS2, WS2, and MoSSe, we achieve volatile,
analog memristive behavior with ultra-low energy consumption and full CMOS compatibility. These devices operate adaptively, mimicking short-term synaptic responses—something oxide-based memristors cannot deliver—and provide the foundation for real-time tasks such as vision, speech recognition, and neuromorphic signal processing. Volatile memristors offer a direct solution to the von Neumann bottleneck by merging memory and computation at the sensor edge. Instead of wasting energy shuttling data back and forth, information is processed where it is generated. This drastically reduces latency and power consumption while enabling compact, scalable architectures capable
of pre-processing massive data streams for AI tasks in real time.

This Proof of Concept will advance our patented demonstrator from TRL3 to TRL5. We will fabricate nanoscale, CMOS-compatible devices, integrate them into programmable preprocessing modules, and validate them with industrial partners. The project carries significant risk, requiring new fabrication strategies, system-level integration, and novel architectures. However, the potential reward is substantial: successful outcomes would open the way to using 2D hybrid memristors in applications beyond the reach of conventional
inorganic compounds.