Technology and electrical characterization of MemFlash cells for neuromorphic applications

The influences of varying charge storage capabilities of EEPROM transistors in a MemFlash configuration are presented. The effect a thinning of the tunnelling oxide has on the retention and the corresponding hysteretic current-voltage curves of MemFlash cells is investigated through measurements as well as simulations. Furthermore, the variation of charging and discharging voltages along with different cycle frequencies is explored in respect to the change in hysteretic behavior. Finally, the influences of changing device parameters on the behavior as artificial synapses were investigated by emulating LTP for different MemFlash cells. Here, we found a strong dependency of the learning rate and the memory capabilities from the tunnel oxide thickness, which allows flexible application in neuromorphic computing schemes.