Memristive systems comply many attractive properties, like the analog memory operation with multilevel programmability and the complex temporal dynamics with an exponential tunability of the switching speed upon a modest linear variation of the driving voltage. Thank to these properties memristors offer a unique possibility to implement brain inspired computing architectures fully at the hardware level. However, the application of memristive devices as hardware building blocks for artificial intelligence usually relies on the direct implementation of biological synaptic functionalities and/or software machine learning protocols. This approach does not take full advantage of the rich nonlinear transfer functions and complex temporal dynamics of the specific ReRAM units.
The PhD candidate will experimentally investigate novel neuromorphic functionalities in memristor junctions fully exploiting the specific dynamical properties of the investigated units. This includes the demonstration of long- and short-term memory operation with tunable memory time constant, and the development of autonomous decision-making units, that can distinguish different types of complex driving signals. First a statistical ensemble of scanning tunneling microscope based memristor junctions will be studied, and later the best performing devices will be embedded in an on-chip environment.
Deep knowledge in nanophysics, long experience in experimental physics