Resistive random-access memory

Resistive random-access memory (ReRAM or RRAM) is a type of non-volatile (NV) random-access (RAM) computer memory that works by changing the resistance across a dielectric solid-state material, often referred to as a memristor. One major advantage of ReRAM over other NVRAM technologies is the ability to scale below 10 nm.

ReRAM bears some similarities to conductive-bridging RAM (CBRAM) and phase-change memory (PCM) in that they change dielectric material properties. CBRAM involves one electrode providing ions that dissolve readily in an electrolyte material, while PCM involves generating sufficient Joule heating to effect amorphous-to-crystalline or crystalline-to-amorphous phase changes. By contrast, ReRAM involves generating defects in a thin oxide layer, known as oxygen vacancies (oxide bond locations where the oxygen has been removed), which can subsequently charge and drift under an electric field. The motion of oxygen ions and vacancies in the oxide would be analogous to the motion of electrons and holes in a semiconductor.

Although ReRAM was initially seen as a replacement technology for flash memory, the cost and performance benefits of ReRAM have not been enough for companies to proceed with the replacement. Apparently, a broad range of materials can be used for ReRAM. However, the discovery^[1] that the popular high-κ gate dielectric HfO₂ can be used as a low-voltage ReRAM has encouraged researchers to investigate more possibilities.

RRAM is the registered trademark name of Sharp Corporation, a Japanese electronic components manufacturer, in some countries, including members of the European Union.^[2]

An energy-efficient chip called NeuRRAM fixes an old design flaw to run large-scale AI algorithms on smaller devices, reaching the same accuracy as digital computers, at least for applications needing only a few million bits of neural state. As NeuRRAM is an analog technology, it suffers from the same analog noise problems that plague other analog semiconductors. While this is a handicap, many neural processors do not need bit-perfect state storage to do useful work.^[3]