Exploring Probabilistic Computing with Magnetic Tunnel Junctions
A team of researchers from Northwestern University, University of Messina, Western Digital Corporation, and Universitat Jaume I have recently published a technical paper titled “Probabilistic computing with voltage-controlled dynamics in magnetic tunnel junctions.” This paper delves into the exciting field of probabilistic (p-) computing, which offers a physics-based approach to tackle computational problems that are challenging for conventional von Neumann computers.
The Concept of Probabilistic Computing
Probabilistic computing involves the development of fast, compact, and energy-efficient probabilistic bits that are capable of solving complex computational problems. In this study, the researchers focus on stochastic magnetic tunnel junctions (MTJs) with low energy barriers. MTJs are proposed as potential candidates for implementing p-bits, where the dwell time in each state is controlled by current. However, this approach presents several challenges, such as the need for precise control of energy barriers across a large number of MTJs and the requirement for an analog control signal.
Introducing Voltage-Controlled Magnetic Anisotropy (VCMA) Effect
To overcome these challenges, the researchers demonstrate an alternative p-bit design based on perpendicular MTJs that utilize the voltage-controlled magnetic anisotropy (VCMA) effect to create the random state of a p-bit on demand. This design ensures that the MTJs remain stable in the absence of voltage, with the VCMA-induced dynamics generating random numbers in less than 10 ns per bit.
Furthermore, the researchers propose a compact method of implementing p-bits by utilizing VC-MTJs without a bias current. They conduct experiments to solve integer factorization problems using bit-streams generated by VC-MTJs, proving the feasibility of their proposed p-bits and the high quality of the generated random numbers.
Impact on the Development of P-Computers
This research has significant implications for the development of p-computers. Firstly, it supports a fully spintronic implementation of a p-bit, which is essential for advancing the field. Additionally, it enables true random number generation at a low cost for ultralow-power and compact p-computers implemented in complementary metal-oxide semiconductor chips.
The full technical paper published in September 2023 can be found here (Yixin Shao et al 2023 Nanotechnology 34 495203).
For further exploration into the exciting world of computing, consider reading the article “The Race Toward Quantum Advantage.” It discusses the ongoing competition between quantum computing and conventional computers and poses the question of when quantum computing will surpass its traditional counterparts.