Abstract
In this work, we experimentally demonstrated the spintronic quaternary true random number generator (QTRNG) utilizing the stochastic switching characteristics of the sub-micrometer-scale L10-FePt spin–orbit torque device. The device exhibits good perpendicular magnetic anisotropy and inherent thermal-fluctuation-induced randomness during magnetization switching. A comparator-based quantization architecture is employed to convert the stochastic analog Hall resistance signals into uniformly distributed quaternary digital outputs, achieving near-ideal randomness with deviations below 2%. Practical implementations explored include a QTRNG-based mosaic masking technique, demonstrating superior visual obfuscation with high Shannon entropy (∼6.88 bits) and convolutional neural networks initialized with QTRNG-derived kernels. This work demonstrates the potential of FePt-based QTRNG systems for reliable multilevel random number generation, establishing a foundation for their integration into advanced spintronic solutions for information security and future deployment in fully spintronic neuromorphic hardware systems.
Funding Information
  • National Natural Science Foundation of China (61674062)
  • National Natural Science Foundation of China (51501168)
  • National Natural Science Foundation of China (41574175)
  • National Natural Science Foundation of China (41204083)
  • Fundamental Research Funds for the Central Universities of the China University of Geosciences (CUG150632)
  • Fundamental Research Funds for the Central Universities of the China University of Geosciences (CUGL160414)

This publication has 12 references indexed in Scilit: