Stack-Engineered Mode Selection in PtMn/(Co/Pd)_n Multilayers Enables Deterministic Analog Spin-Orbit Torque Synapses

Area of Science:

• Materials Science
• Spintronics
• Neuromorphic Computing

Background:

• Spin-orbit-torque (SOT) devices offer potential for both memory and neuromorphic computing.
• Controlling the switching mode (binary vs. analog) is crucial for device applications.
• PtMn/(Co/Pd)n multilayers are investigated for their SOT switching characteristics.

Purpose of the Study:

• To demonstrate that the Co/Pd repeat number (n_Co/Pd) in PtMn/(Co/Pd)n multilayers dictates the magnetization reversal mechanism and switching mode.
• To explore the potential of these multilayers as energy-efficient analog SOT synapses for neuromorphic hardware.

Main Methods:

• Fabrication of PtMn/(Co/Pd)n multilayers with varying n_Co/Pd.
• Investigation of magnetization reversal mechanisms (nucleation, domain-wall propagation) via material parameter tuning.
• Electrical conditioning of SOT devices to induce switching mode transitions.
• Evaluation of device performance for neuromorphic applications, including synaptic plasticity and classification accuracy.

Main Results:

• For n_Co/Pd ≤ 7, binary switching via nucleation and domain-wall propagation was observed, transitioning to analog behavior with electrical conditioning.
• For n_Co/Pd ≥ 8, suppressed domain-wall propagation led to intrinsic nucleation-dominated analog switching.
• High-n_Co/Pd stacks exhibited enhanced analog behavior (smoother potentiation/depression, more intermediate states) and >97% neuromorphic classification accuracy after current conditioning.

Conclusions:

• The Co/Pd repeat number is a key material parameter for controlling SOT switching modes in PtMn/(Co/Pd)n multilayers.
• Stack design and hybrid tuning offer scalable strategies for developing energy-efficient analog SOT synapses.
• These findings advance the integration of spintronic devices into neuromorphic computing architectures.