Superparamagnetic and Stochastic-Write Magnetic Tunnel Junctions for High-Speed True Random Number Generation in Advanced Computing
Jonathan Z. Sun, Christopher Safranski, Siyuranga Koswata, Pouya Hashemi, Andrew D. Kent · 2025
Read Superparamagnetic and Stochastic-Write Magnetic Tunnel Junctions for High-Speed True Random Number Generation in Advanced Computing by Jonathan Z. Sun, Christopher Safranski, Siyuranga Koswata, Pouya Hashemi, Andrew D. Kent (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-13469.txt.
Introduction
Jonathan Z. Sun and 4 co-authors review two chip-scale ways to make true random numbers in a 2025 paper. True random bits are a must for codes and secure computing, and small, low-power sources fit best on chips. Both ways use a magnetic tunnel junction, a tiny part whose resistance flips with its magnetic state.
One way lets heat noise flip a free magnetic layer on its own, and reads the stream at high speed with low power. The other uses a stable layer and fires write pulses that flip it about half the time. That also yields a fast random stream and fits standard chip flows used for magnetic memory. Both passed the standard national test suites for randomness. Both also take up a tiny area and use far less energy than older sources.
Catalog record
| CREATOR | Jonathan Z. Sun, Christopher Safranski, Siyuranga Koswata, Pouya Hashemi, Andrew D. Kent |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 41 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-13469
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Superparamagnetic and Stochastic-Write Magnetic Tunnel Junctions for High-Speed True Random Number Generation in Advanced Computing Jonathan Z. Sun, Christopher Safranski, Siyuranga Koswata, Pouya Hashemi, Andrew D. Kent Abstract We review two magnetic tunnel junction (MTJ) approaches for compact, low-power, CMOS-integrated true random number generation (TRNG). The first employs passive-read, easy-plane superparamagnetic MTJs (sMTJs) that generate thermal-fluctuation-driven bit streams at 0.5–1 Gb/s per device. The second uses MTJs with magnetically stable free layers, operated with stochastic write pulses to achieve switching probabilities of about 0.5 ( i.e. , write error rates of ≃ 0.5 \simeq 0.5 ), achieving ≳ 0.1 \gtrsim 0.1 Gb/s per device; we refer to these as stochastic-write MTJs (SW-MTJs). Randomness from both approaches has been validated using the NIST SP800 test suites. sMTJ approach uses a read-only cell with low power and can be compatible with most advanced CMOS nodes, while SW-MTJs leverage standard CMOS MTJ process flows, enabling co-integration with embedded spin-transfer torque magnetic random access memory (STT-MRAM). Both approaches can achieve deep sub-0.01 μ \mu m 2 MTJ footprints and offer orders-of-magnitude better energy efficiency than CPU/GPU-based generators, enabling placement near logic for high-throughput random bit-streams for probabilistic computing, statistical modeling, and cryptography. In terms of performance, sMTJs generally suit applications requiring very high data-rate random bits near logic processors, such as probabilistic computing or large-scale statistical modeling. Whereas SW-MTJs are attractive option for edge-oriented microcontrollers, providing entropy sources for computing or cryptographic enhancement. We highlight the strengths, limitations, and integration challenges of each approach, emphasizing the need to reduce device-to-device variability in sMTJs—particularly by mitigating magnetostriction-induced in-plane anisotropy—and to improve temporal stability in SW-MTJs for robust, large-scale deployment. 1 Introduction A nanoscale thin-film magnetic tunnel junction (MTJ) is the central element in spin-transfer-torque magnetic random access memory (STT-MRAM) offered by advanced CMOS foundries Coughlin 2023 ; Coughlin 2024 ; Naik et al. 2021 ; Chen et al. 2021 ; Kent and Worledge 2015 . In a similar technology setting, but with modified device structures, circuit designs, and operating modes, the same MTJ can serve as a compact, low-power, back-end-of-line (BEOL) CMOS-integrated source of entropy for advanced computing Yang et al. 2018 ; Camsari et al. 2017a ; Camsari et al. 2019a ; Camsari et al. 2019b ; Kaiser et al. 2020 ; Kaiser and Datta 2021 ; Aadit et al. 2022 ; Smith et al. 2022 ; Chowdhury et al. 2023 ; Suresh and Burleson 2010 ; Cao et al. 2022 . In this review, we examine the key attributes, performance metrics, and integration challenges of MTJ-based entropy sources for advanced computing applications. STT-MRAM is primarily used as an embedded memory in which perpendicularly magnetized MTJs (pMTJs) Worledge et al. 2011 ; Ikeda et al. 2010 are closely integrated directly above logic circuits, enabling data rates well matched to logic operations Edwards et al. 2020 ; Hu et al. 2022 . Using similar structures, stochastic MTJs can serve as fast (gigabit per second per MTJ – Gbps/MTJ, or higher), high-density ( \ll 100 nm MTJ diameter), and low-power random bit generators that are BEOL integrated with logic for workloads requiring large volumes of random data. Such high-quality random bit streams may also help hardware implementations of quantum-safe cryptography. For these applications, the viability of an MTJ-based approach depends on several key attributes: (1) the speed of random bit-stream generation, (2) the statistical quality of the output sequence, and (3) the consistency and ease with which the bit-stream properties can be controlled. Figure 1: Overview of MTJ-based entropy sources, comparing superparamagnetic fluctuation devices (sMTJs), their subtypes, and stochastic write devices (SW-MTJs), integration compatibility, and typical application domains. Figure 1 shows the classes of MTJ-based entropy sources. Two leading approaches are: (1) MTJs operating in the superparamagnetic regime (sMTJs), in which thermal agitation drives magnetic fluctuations of a MTJ magnetic free layer, and (2) Magnetically stable MTJs operated at reduced write voltage or pulse duration to intentionally produce probabilistic switching characteristics, referred to as a stochastic write MTJ (SW-MTJ). For superparamagnetic MTJs, two general categories can be distinguished, as shown on the left-hand side of Fig. 1 . The first, referred to here as uniaxial sMTJs, employs a low-barrier uniaxial anisotropy MTJ in which the free-layer moment—and thus the tunnel magnetoresistance—fluctuates between two well-defined states: parallel and antiparallel, corresponding to low- and high-resistance levels. The second, referred to as easy-plane sMTJs, employs an MTJ with strong easy-plane magnetic anisotropy, confining thermal fluctuations of the magnetic moment essentially to the film plane. In this case, at any instant, ideally, the moment has equal probability of pointing in any in-plane direction, leading to a tunnel magnetoresistance as analog white-noise signal up to a cut-off frequency f_{\mathrm{c}}\sim\gamma H_{\mathrm{p}} associated with the easy-plane anisotropy field H_{p}\sim 4\pi M_{s} where M_{s} is the saturation magnetization of the free-layer (FL) of the MTJ Aron et al. 2014 . This analog resistance-fluctuation signal can then be digitized or otherwise processed to generate a random bit stream. These two approaches — the read-only, thermal-fluctuation based superparamagnetic MTJ (sMTJ) and the stochastic write MTJ (SW-MTJ) — as illustrated in Fig. 1 , are suited for different applications and have different technology-integration attributes. The sMTJ approach operates by passively reading the device resistance to obtain entropy, consuming minimal power. Single-device demonstrations have achieved bit rates of 0.5–1 Gb/s per sMTJ. Because it does not require a current-intensive write operation, the sMTJ is compatible with the most advanced CMOS nodes, provided BEOL MTJ fabrication can be optimized for consistent performance metrics (discussed below). This makes it particularly suitable for circuits and systems that demand the highest possible random bit rate at minimal power, such as large-scale probabilistic computing Aifer et al. 2025 ; Nikhar et al. 2024 ; Singh et al. 2024 ; Bunaiyan et al. 2023 . The SW-MTJ approach operates by actively strobing the MTJ with write pulses, making it inherently more drive-current and power intensive and somewhat slower ( \lesssim 10\times ) in maximum achievable bit rate than the sMTJ. Its main advantage is that it requires only minor circuit modifications relative to standard STT-MRAM technology, often using the same pMTJ devices (or slightly modified versions, e.g. , medium energy barrier free layers). The same pMTJ process can be shared between STT-MRAM and SW-MTJ-based probabilistic bit (p-bit) circuits, enabling both to coexist on the same chip. This makes the SW-MTJ well suited for applications such as integrated cryptographic functions that frequently require substantial volumes of random bits but do not demand the highest possible bit rate. Below, we explore in greater depth the distinctive characteristics of each approach. We review the underlying device physics and current technological status, and identify key materials, devices, and engineering challenges that must be addressed to advance these MTJ-based entropy sources toward large-scale deployment. 2 Superparamagnetic MTJ as entropy sources A superparamagnetic magnetic tunnel junction (sMTJ) is an MTJ in which the variation in magnetic anisotropy energy of the free-layer magnetization with orientation is comparable to or smaller than thermal energy k_{B}T . The magnetization can thus fluctuate over a wide range of angles, producing tunnel magnetoresistance variations spanning the full range between the parallel and antiparallel states, as conceptually illustrated in Fig. 2 . The specific form of