Altermagnets, as an emerging type of magnet, integrate the key advantages of both ferromagnets and antiferromagnets, as they possess a spin-splitting band structure similar to ferromagnets, and maintain a vanishing net magnetic moment as in antiferromagnets. These features make the altermagnets promising candidates for high-speed, high-density, robust, and easily readable memory devices. External magnetic field and spin transfer torque, generated by charge current flowing through the magnetic tunneling junction, have been successively adopted to switch between the binary states “0” and “1” in magnetic random access memory devices by the industry. Spin-orbit torque (SOT), which is expected to further reduce the writing-current density and enhance endurance, has been regarded as a key write technology for next-generation magnetic random access memory. However, theoretical understanding of SOT-induced magnetization dynamics in altermagnets remains largely unexplored.

In this paper, the SOT-induced magnetization dynamics in altermagnets are investigated with the help of atomistic spin simulation performed with VAMPIRE. Both the dynamical evolution process during the action of SOT and the post-pulse relaxation are analyzed. During SOT application, altermagnets exhibit oscillation behaviors very similar to those of antiferromagnets, but once the damping-like SOT field exceeds the critical threshold, the switching time will be significantly shortened. In the relaxation stage, the altermagnetic sublattice moments display distinct chaotic dynamical characteristics—evidenced by the positive maximum Lyapunov exponent—which do not exist in antiferromagnets. Parameter-sweep simulations further indicate that the chaotic behavior is strongly influenced by the anisotropic exchange coupling (a defining feature of altermagnetism) and the Gilbert damping parameter. Moreover, when subjected to periodic SOT pulses, altermagnet exhibits random switching between multiple stable states, which is likely to be determined by anisotropic exchange coupling rather than magnetocrystalline anisotropy. This stochastic multi-state response indicates that altermagnets may serve as useful building blocks for probabilistic computing, neuromorphic computing, and random logic.

In summary, our results reveal the following three key findings:

1) Altermagnets may switch faster than antiferromagnets under SOT, indicating higher memory-writing speed.

2) Relaxation dynamics with chaotic behavior introduce new physical dimensions for probing and exploiting altermagnetism.

3) Multi-pulse-induced stochastic switching demonstrates potential applications of altermagnets in computing-in-memory architectures and nontraditional computing paradigms.