Magnetic skyrmions are appealing for use in logic and memory devices combining small size and fast motion. Here, we propose to exploit skyrmion interactions to perform both logic and memory operations at the nanoscale. Our concept opens a path for novel devices which intrinsically merge high-density non-volatile data storage with computing capabilities.
a. Schematic representation of the racetrack memory. b. Simulation of a racetrack device where skyrmions are stored into boxes (region of lower magnetic anisotropy). c. Simulation of a magnetic full adder. d. Device combining racetrack and logic devices.
Magnetic skyrmions are currently fascinating many research groups in the world, as they could offer a new way to store and process information in our computers. These nanoscale magnetic textures are composed of elementary nanomagnets that wind up to form a stable spiral structure, like a well tighten node. Skyrmions can be manipulated by very low electrical currents, which opens a path for their use as information carriers in computing devices. Several groundbreaking memory and logic devices have been proposed, that promise large information density and low power consumption. A prime example is the racetrack memory where trains of skyrmions in a track are moved toward the read and write elements using current pulses (see Figure1a). Here the information is encoded in the position of the skyrmion in the train, where for instance, the presence of one skyrmion means 1, while no skyrmion means 0. However, this device is prone to error: if one skyrmion goes faster or slower than the others or gets stucks by some defects in the materials, it may move forward or backward in the train, and change the information encoded. To solve this issue, we proposed to define some boxes in the track where the skyrmions get trapped (see Figure 1B). This is performed by locally modifying the magnetic properties of the material (here magnetic anisotropy), using light (He+) ion irradiation for instance. Simulations show that the skyrmions are indeed stable in the box and can be moved synchronously and reliably between the boxes using current pulses. More complex shapes can also be defined to perform logic operations. For instance, in the device shown in figure 1C, a larger box is defined where the skyrmions interact: the natural repulsion between the skyrmions in the box leads to a deviation of their trajectory and allows to perform a full adder operation. These gates can be cascaded to perform simple logic operations such as AND, OR, NOT, NAND, XOR, and NXOR. Both racetrack memory device and logic gates can also be combined (see Figure 1d) such that memory and logic operations are performed at the same place. The integration of such a device in our computer would allow a dramatic decrease of their consumption, since it would save the high energy cost and time required to move data between the memory and computing units.
Further reading: Robust and Programmable Logic-In-Memory Devices Exploiting Skyrmion Confinement and Channeling Using Local Energy Barriers, N. Sisodia, J. Pelloux-Prayer, L. Buda-Prejbeanu, L. Anghel, G. Gaudin, O. Boulle, Phys. Rev. Appl., 18, 014025 (2022). Open access: hal-03740728v1
Contacts: Olivier BOULLE