Figure 5: Time and temperature dependence of the resistance levels and noise characteristics.
(a) The programmed states in a phase-change device without projection component exhibit significant temporal resistance drift at constant temperature with a drift coefficient, νR, determined by a fit to a power law (black line), of ≈0.067 that is characteristic of the AIST phase-change material. (b) The programmed states in a projected memory device exhibit strongly reduced temporal resistance drift at constant ambient temperature. The black lines are fits to a power law. (c) Applied temperature to investigate the temperature dependence of the programmed states. (d) Resistance variations corresponding to two different programmed states in a non-projected memory device when the ambient temperature is varied as shown in c. (e) Resistance variations corresponding to five different programmed states in a projected memory device when the ambient temperature is varied as shown in c. The black lines show the resistance variation predicted by FE simulations. (f) Normalized spectral density of the current noise for both a projected resistive memory device and an identical phase-change device without projection component. The measurements are obtained for three different read voltages (VR). The projected memory cell is programmed to a resistance of 24 kΩ and the non-projected cell is programmed to 1.1 MΩ, both corresponding to amorphous regions of about 90 nm in length. The dashed lines show the thermal noise floors for each bias voltage and each resistance.
