- With the advent of mobile and handheld electronic devices, the demand for much smaller, faster and ultra-low power systems keeps growing. Yet to meet such needs, the microelectronics industry cannot rely anymore on following Moore’s law like it has for the last decades.
- Embedded memories, which represent a major part of the circuits silicon area have now become a major contributor to power dissipation in integrated system circuits. To solve these issues, several technologies are intensively investigated to replace existing embedded memories (SRAM and Flash).
- Spin-Transfer Torque Magnetic Random Access Memory (STT-MRAM) has been chosen by the industry as the non-volatile memory technology of choice to replace Embedded Flash at advanced technology nodes.
Memory is ubiquitous and fundamental to all computing systems, yet incumbent technologies are still dominant, mostly unchanged for decades but facing increasingly higher hurdles placed by Moore’s Law. SOT’s unique and revolutionary attributes combine performance and reliability requirements for both RAM and NVM applications.
- Fast enough to match logic speed
- High endurance for reliable cache/buffering
- Easy to process for low cost on-chip integration
What is SOT-MRAM?
Spin Orbit Torque Magnetic Random-Access Memory (SOT-MRAM) is the latest generation of MRAM. Unlike STT-MRAM, SOT offers unique system-level value thanks to its unique combination of high working speed and truly infinite endurance.
Advantages of SOT-MRAM
The absence of high-voltage stress on the device allows for practically infinite SOT-MRAM endurance even at the fastest, sub-ns write speeds. Furthermore, SOT-switching does not need to rely on thermal activation to initiate switching, which makes reliable sub-ns switching with no incubation delays inherently feasible. As a result, SOT-MRAM may be used as DRAM-like working memory or SRAM-like cache memory, which is impossible with STT-MRAM.
MAGNETIC TUNNEL JUNCTION (MTJ)
A magnetic tunnel junction is composed of two ferromagnetic layers separated by a thin insulating (tunnel) barrier. The MTJ stores the binary memory value as the magnetization direction of one of the ferromagnetic thin films, which is called the “Free Layer”. The other ferromagnetic layer is called the “Fixed Layer”(or “Reference Layer”) for its magnetization never moves. The electric resistance of the MTJ is larger when the magnetizations of the two ferromagnetic layers are antiparallel (coding a “0”), while it is lower when the two magnetizations are parallel (coding a “1”). This effect is called tunnel magnetoresistance and is used for reading the data. Writing consists of switching the magnetization of the free layer from one direction to the other, which can be done by a magnetic field, STT or SOT.
What is SOT ?
Two physical phenomena are understood to be at the origin of the spin-orbit torques (SOT): A bulk component, the Spin Hall Effect, and an interfacial component, commonly known as the Rashba Effect.
The Spin Hall effect induces a spin current transverse to the charge current flowing in the SOT layer, leading to a spin accumulation at the SOT interfaces, which then diffuses into magnetic materials.
The Rashba effect originates from the uncompensated electric field at the interface, resulting in an effective magnetic field directly acting on nearby magnetizations.
Both effects lead to magnetic torques, that can control the magnetization direction of an adjacent magnetic “storage” layer, to write a binary “0” or “1” in the MRAM cell.