Unconvetional Computing and Quantum Sensing

Contact: Alberto Riminucci

One of the basic functional anatomical features of the brain is the synapse, a small area of a neuron's receiving dendrites that connects it to another neuron's outgoing axon; the synapse is believed to be one of the building blocks of the brain's intelligence. Simulating synapses a conventional computer program means having to deal with the memory access bottleneck that slows down performance and increases the energy budget. In order to achieve performances comparable to the brain's, especially in terms of energy efficiency, it is necessary to resort to hardware neuromorphic implementations. The foremost approach in solid-state physics consists in using resistive switching devices in which the strength of the synapse is embodied in their conductance. In our laboratory we use the unique combination of a molecular semiconductor and a valence change memory material (AlOx) comprised between two ferromagnetic electrodes, thus forming a molecular spin valve. In this way, we add a second means to control the conductance of the synaptic devices, that is the magnetic field. The molecular spin valve shows both conventional resistive switching and magnetoresistance: by sweeping an applied magnetic field, the magnetizations of its ferromagnetic electrodes can be set either parallel, corresponding to a low conductance state, or antiparallel, corresponding to a high conductance state. The addition of this degree of freedom speeds up the learning rate in a simulation based on a reinforcement learning algorithm.

Selected Publications

Prezioso, Mirko, et al. "A single‐device universal logic gate based on a magnetically enhanced memristor." Advanced Materials 25.4 (2013): 534-538.

Shumilin, Andrei, et al. "Glassy Synaptic Time Dynamics in Molecular La0. 7Sr0. 3MnO3/Gaq3/AlOx/Co Spintronic Crossbar Devices." Advanced Electronic Materials (2024): 2300887.

Bergenti, Ilaria, et al. "Oxygen impurities link bistability and magnetoresistance in organic spin valves." ACS applied materials & interfaces 10.9 (2018): 8132-8140.