An easy-plane anisotropy dominant stochastic magnetic tunnel junction as a circuit entropy source

Abstract

We explorea nanostructured stochastic magnetic tunnel junction in its superparamagnetic limit as asub-nanosecond-speed entropy source for direct back-end integration into CMOS-logic circuits. The minimum autocorrelation time for the magnetoresistance signal from an idealized macrospin superparamagnetic magnetic tunnel junction(SMTJ)is governed by magneto-thermodynamic fluctuations. Earlytheoretical1and numerical studies2,3indicate that, in a simple uniaxial anisotropy potential below thermal energy𝑘𝐵𝑇, the temporal correlation time has a lower limit of around 10~20 ns for conventional magnetic materials-based structures, and that ina combined strong easy-plane and weak in-plane uniaxial anisotropy potential≪𝑘𝐵𝑇, a nanomagnet can fluctuate faster, with sub-nanosecond auto-correlation timethatin the simplest limits cales with √𝑉𝑘𝐵𝑇⁄, assuming full demagnetization easy-plane field of 4𝜋𝑀𝑠and with 𝑉as the macrospin nanomagnet’s volume. Ittherefore seemed possible to use superparamagnetic nano-magnetic tunnel junctions for passive sub-nanosecond stochastic signal generation. Recent experimental works from other groups4,5as well as ours6indicatethe existence of such high-speed thermal fluctuation processes in SMTJs. In this talk, we will discuss the baseline experimental findings with SMTJs made using similar materials and processes as for spin-torque magnetic random-access memory (STT-MRAM) that are ready for CMOS back-end integration7,8. We give a specific set of test and characterization protocols for device-level metrologies and discuss several materials and device-physics related factors one could further investigate and optimize. Those include the role of a combined strong easy-plane and a weak in-plane anisotropy on fluctuation, the combined stochastic dynamics of the free and reference layer of such magnetic tunnel junctions under bias-voltage induced spin-current drive, and the dynamics’ dependence on bias voltage, on magnetic field, and on device and materials parameters in need of control. These provide a baseline for future applications-specific optimization.To characterize the randomness of the resulting fluctuation signalbeyond its auto-correlation time, the SMTJs signal is binary digitized with varying sampling speed,up to 0.25ns, andthe resulting bit-streambenchmarked using the NIST-SP800-22r1a test suite. Single-device bit-stream is shown to pass tests withbitrate around 250Mb/sec. With two bit-stream XORed, passingbit ratesofgreater than 1Gb/sec is seen. With these measurements, wehaveidentified several device materials and fabrication factors to optimize for faster performance limits, and for improving device-to-device uniformity –factors that are important for advanced un-conventional computing and cryptography. 1William Fuller Brown, “Thermal fluctuation of a single-domain particle”, Phys. Rev. 130, 1677 (1963). 2Orchi Hassan et al., “Low barrier magnet design for efficient hardware binary stochastic neurons”, IEEE Magn. Lett. 10,4502805 (2019). 3J. Kaiser et al., “Subnanosecond fluctuations in low-barrier nanomagnets”, Phys. Rev. Appl. 12, 054056 (2019). 4K. Hayakawa et. al, “Nanosecond random telegraph noise in in-plane magnetic tunnel junctions”, Phys. Rev. Lett. 126, 117202 (2021). 5L. Schnitzspan et al., “Nanosecond true-random-number generation with superparamagnetic tunnel junctions: identification of Joule heating and spin-transfer-torque effects”, Phys. Rev. Appl. 20, 024002 (2023). 6C. Safranski et al., “Demonstration of nanosecond operation in stochastic magnetic tunnel junctions”, Nano Letters, 21,2040 (2021). 7J. Z. Sun, et al., “Stochastic magnetic tunnel junction with easy-plane dominant anisotropy”, Phys. Rev. B 107, 184433(2023).8J. Z. Sun, et al, “Easy-plane dominant stochastic magnetic tunnel junction with synthetic antiferromagnetic layers”, Phys. Rev. B 108, 064418 (2023)