Other meanings of Racetrack memory
Computer Science
Racetrack memory is a proposed non-volatile computer memory technology that stores data in magnetic domain walls along nanoscopic wires, offering the potential for high density, fast access, and low power consumption. It was first proposed by Stuart Parkin at IBM Almaden in 2002, and its name derives from the way data bits race around a track-like structure.
Racetrack memory stores data in the magnetic polarization of domain walls—boundaries between regions of opposite magnetization—along a ferromagnetic nanowire. Data is written by a local magnetic field or spin-polarized current, and read by a magnetic tunnel junction (MTJ) positioned near the wire. To access a particular bit, a current pulse moves the entire pattern of domain walls along the wire, shifting the desired bit to the read/write head. This shift mechanism is analogous to a shift register, but with magnetic rather than electronic storage.
The key innovation is the use of spin-transfer torque (STT) to move domain walls, which was demonstrated experimentally in 2004 by Parkin's group at IBM. Unlike conventional magnetic storage, which requires mechanical motion, racetrack memory is entirely solid-state, enabling faster access and higher endurance.
The primary advantage of racetrack memory is its potential for extremely high storage density, because the nanowires can be arranged vertically, forming a 3D array. This could achieve densities exceeding 1 terabit per square inch, surpassing flash memory. Additionally, it offers non-volatility, high speed (comparable to DRAM or SRAM), and low power consumption, as no power is needed to maintain data.
However, several challenges remain. Moving domain walls reliably at high speeds requires precise control of current pulses and material properties. Defects and imperfections in the nanowire can pin domain walls, causing errors. Also, the read/write head must be positioned extremely close to the wire, which is difficult in a 3D architecture. Research has focused on materials like permalloy and cobalt, and on techniques such as using notches or engineered defects to control domain wall motion1.
The concept of racetrack memory was introduced by Stuart Parkin in 2002, and a prototype was demonstrated in 2008 by IBM, showing that data could be shifted along a nanowire using current pulses. Since then, research has progressed in universities and industry, with notable contributions from groups at the University of California, Berkeley, and the University of Leeds.
In 2015, a team at IBM demonstrated a three-dimensional racetrack memory device, confirming the feasibility of vertical stacking2. However, as of 2024, racetrack memory has not been commercialized, and it remains an active area of research, with recent work focusing on using antiferromagnetic materials and skyrmions as alternatives to domain walls.
Beyond the mainstream narrative, several niche aspects are noteworthy. One is the use of racetrack memory in extreme environments, such as space, where its radiation hardness and non-volatility are advantageous. Another is the exploration of domain wall motion driven by spin-orbit torque, which could reduce the current required for shifting data.
An overlooked figure is the physicist Luc Thomas, who co-authored key papers with Parkin and contributed to the development of the 3D prototype. Additionally, the concept of using racetrack memory for neuromorphic computing has been proposed, where the shift register nature could mimic synaptic weights.
Trivia: The name "racetrack" was inspired by the circular track used in early experiments, and the technology is sometimes called "domain wall memory" in the literature.
This article focuses on the proposed non-volatile memory technology using magnetic domain walls, not the racetrack memory concept in other contexts.
Help improve the encyclopedia. Reports go straight to the site manager.