Reported in Nature Scientific Reports as ‘Multibit memory operation of metal-oxide bi-layer memristors‘, the researchers reached 128 levels by evaluating several configurations of functional oxide materials.
“Memristors are a key enabling technology for next-generation chips, which need to be highly reconfigurable yet affordable, scalable and energy-efficient,” said Professor Themis Prodromakis. “This technology is ideal for developing novel hardware that can learn and adapt autonomously, much like the human brain.”
The research compared a reference device was 20×20μm with platinum top and bottom electrodes, separated by a TiO2 layer, with experimental devices using the same electrode arrangement, but separated by bi-layers of: AlxOy/TiO2, TaxOy/TiO2, SiO2/TiO2, ZnO/TiO2, HfOx/TiO2 or WOx/TiO2. In each case there was 40nm of TiO2 and 4nm for the other oxide.
Variability over time meant TiO2-only devices could only store 10 resistive states,
According to the paper, AlxOy/TiO2 was the best, with 92 states (6.5 bits) on a single cell at the time the paper was sent for publishing. Retention was demonstrated over 8hr at both room temperature and 85°C.
Data was stored by first applying a series of 100ns 2V pulses to push cell to a ground resistance state (~27kΩ), then using 100ns pulses of the other polarity to step resistance up to predictable higher resistances up to ~80kΩ.
The work was supported by the Engineering and Physical Sciences Research Council (EPSRC) and the Royal Society.
Prodromakis and colleagues will present seven research papers at ISCAS 2018 in Florence in May.