Answers are grounded in this paper and cite their sources.
Physical information processing relies on the ability of a dynamical system to transform external stimuli into distinguishable internal states. Although nonlinear magnetization dynamics in magnetic materials exhibit memory and rich dynamical behavior, engineering responses that depend on the temporal ordering of inputs remains a central challenge for spintronics-based physical computing. Here, we show that exchange-biased PtMn/NiFe heterostructures create input-selective magnetic state spaces through history-dependent magnetization dynamics driven by microwave excitation. Above a threshold microwave power, nonlinear excitation modifies the interfacial magnetic configuration so that sequences delivering identical total excitation energy but differing in temporal order evolve into distinct resonance states. Short four-bit microwave pulse sequences generate sixteen experimentally distinguishable magnetic resonance conditions, with only a selective subset of patterns producing large state transitions and yielding sparse activation. Using the shifted resonance fields as physical features, a multinomial logistic regression classifier achieves 88.5% accuracy on the Modified National Institute of Standards and Technology (MNIST) handwritten-digit recognition task. These results establish exchange-biased antiferromagnet/ferromagnet interfaces as an effective platform for engineering history-dependent magnetic state spaces and demonstrate that the interfacial spin dynamics can perform temporal information processing of short pulse sequences directly in hardware, with relevance to edge artificial intelligence, radar, and sensing applications.
No takes yet. Share an insight, caveat, or question.
Mohan et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: