Description:

Probabilistic Ising machines (PIMs) can potentially solve many computationally hard problems more efficiently than deterministic algorithms on von Neumann computers. Stochastic magnetic tunnel junctions (S-MTJs) are potential entropy sources for PIMs. However, scaling up S-MTJ PIMs requires fine control of a small magnetic energy barrier and duplication of area-intensive digital-to-analog converter (DAC) elements across large numbers of devices. Here we discuss the design of digital probabilistic computers that utilize two classes of unconventional magnetic tunnel junctions, which can overcome many of these limitations: (1) Voltage-controlled magnetic tunnel junctions (V-MTJs) [1-3] and (2) All-antiferromagnetic tunnel junctions (ATJs) [4, 5]. 

We first report a probabilistic computer that is based on an application-specific integrated circuit (ASIC) fabricated in 130-nm foundry CMOS and uses V-MTJs as its entropy source [3]. With this system, we implement integer factorization as a representative hard optimization problem using invertible logic gates created with 1,143 probabilistic bits. The ASIC uses stochastic bit sequences read from an adjacent V-MTJ chip. The V-MTJs are thermally stable in the absence of voltage and synchronously generate random bits without the use of DAC elements, using the voltage-controlled magnetic anisotropy (VCMA) effect.

Next, we present recent experimental results on ATJs based on noncollinear antiferromagnetic PtMn3, where we have demonstrated record-high room-temperature tunneling magnetoresistance ratios (>300%) along with electrical current-induced switching of the noncollinear magnetic (octupole) order [4, 5]. We identify a new current-induced torque mechanism – octupole-driven spin-transfer torque (OTT) – which allows for electrical switching of these ATJs, and discuss their prospects for PIM, memory, and THz electronics applications.

Pedram Khalili, Professor of Electrical and Computer Engineering, Northwestern University

Host: Istvan Kovacs