Magnetic random access memory (MRAM) could change the way electronic devices store and process data by coding information with magnetic orientations instead of electric charge. The technology uses parallel and anti-parallel states to represent ones and zeroes and is intrinsically nonvolatile, meaning it can keep information without being electrically powered.

However, nonvolatile memory often comes with a trade-off between the retention and the writing of information. A higher energy barrier between parallel and anti-parallel states results in more stability and retention but more difficult switching and writing.

Hazen et al. managed to improve both aspects of a MRAM device, despite their apparent contradiction.

Usually, MRAM technology has two layers: a bottom layer of fixed magnetization and a second storage layer with switchable states. As electrons are injected into the storage layer, they exert a spin transfer torque which forces the storage magnetization to follow suit.

“We added an assistance layer, a second polarizing layer, which is above the storage layer,” said author Bernard Dieny. “So, now we have a third layer on top, which we refer to as the assistance layer, that is going to provide a second spin transfer torque influence on the storage layer.”

The assistance layer has a switchable magnetization and always runs parallel to storage layer after switching. It reinforces the thermal stability and reduces the write current. Compared to similar multilayer devices, it requires fewer layers and is easier to etch.

The team conducted real time measurements of the device resistance and demonstrated how the layers switch dynamically. They aim to improve the constituent materials in the future.

Source: “Real-time investigation of double magnetic tunnel junction with a switchable Assistance layer for high efficiency STT-MRAM,” by Daniel Sanchez Hazen, Bruno Miguel da Silva Teixeira, David Salomoni, Stephane Auffret, Laurent Vila, Ricardo C. Sousa, Ioan Lucian Prejbeanu, Liliana D. Buda-Prejbeanu, and Bernard Dieny, APL Materials (2022). The article can be accessed at

This paper is part of the Materials Challenges for Nonvolatile Memory Collection, learn more here.