Scientists from Tohoku University pioneered the technology of nanosecond operation of the probabilistic bit based on spintronics (p-bit), called the “poor man’s quantum bit” (q-bit).
The late physicist R.P. Feynman introduced a probabilistic computer capable of working with probabilities at scale to provide efficient computation. Using spintronics, the latest technology took the first step towards realizing Feynman’s vision.
Magnetic Tunnel Junction (MTJ) is a key component of nonvolatile memory, or MRAM, a mass storage technology that uses magnetization to store information. There, thermal fluctuations usually pose a threat to the stable storage of information.
P-bits, on the other hand, deal with these thermal fluctuations in thermally unstable (stochastic) magnetic tunnel junctions. Previous collaborative research between Tohoku University and Purdue University has demonstrated a spintronics-based probabilistic computer at room temperature, consisting of stochastic magnetic tunnel junctions with millisecond relaxation times.
To make probabilistic computers a viable technology, it is necessary to develop stochastic magnetic tunnel junctions with much shorter relaxation times that reduce the timescale of p-bit fluctuations. This will effectively improve the speed and accuracy of the calculations.
The researchers have created a nanoscale magnetic tunnel junction device with a flat magnetic axis. The magnetization direction is updated every 8 nanoseconds on average – 100 times faster than the previous world record.
The group explained the mechanism of this extremely short relaxation time in terms of entropy, a physical quantity used to represent the stochasticity of systems not previously considered for magnetization dynamics. Deriving a universal equation governing the entropy in the dynamics of magnetization, they found that the entropy increases rapidly in magnetic tunnel junctions with an in-plane easy axis with large values of perpendicular magnetic anisotropy. The group deliberately used a flat magnetic easy axis to achieve shorter relaxation times.
The developed MTJ is compatible with current semiconductor processing processes and shows significant prospects for the future implementation of high-performance probabilistic computers. This theoretical basis for the dynamics of magnetization, including entropy, is also of broad scientific importance, ultimately showing the potential of spintronics in solving controversial issues in statistical physics.