Voltage Scaling in Area Scalable Selector-Less PrMnO3 RRAM by N2: O2 Partial Pressure Dependent Annealing

Abstract

Manganite based Resistance Random Access Memories (RRAM) are attractive as they are non-volatile, area-scalable, with fast switching and large endurance and retention. However, the cross-point RRAM array implementation has sneak path current leakage. To avoid this issue, a selector device with high nonlinearity is used in series with a memory device. This addition of the selector device leads to a higher voltage requirement. Further, the integration of memory and selector is challenging and leads to higher circuit complexity. To resolve this issue, a single device is extensively explored to work as a selector-less RRAM. Recently, we have experimentally demonstrated a nanoscale PrMnO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</sub> (PMO) based selector-less RRAM with record memory window (MW) and non-linearity (NL). Along with MW and NL, the devices should operate at lower voltages to reduce power consumption in large memory array. To address this issue, in this paper, first, we experimentally demonstrate the switching voltage scaling (40 % reduction) in PMO selector-less RRAM using N <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> :O <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> partial pressure dependent annealing in the fabrication process. Further, this voltage reduction is attributed to the lowering of trap density in as-fabricated devices with higher O <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> concentration annealing. Finally, the trap density is extracted which shows the dependence where reduced trap density leads to lower operational voltages. This experimental voltage scaling of selector-less RRAM will provide a significant power performance improvement in the large cross-point memory array and neuromorphic computing applications.