Sub-10 nm HfZrO ferroelectric synapse with multiple layers and different ratios for neuromorphic computing

Area of Science:

• Materials Science
• Computer Engineering
• Electrical Engineering

Background:

• Emerging non-volatile memories are crucial for overcoming the von Neumann bottleneck in neuromorphic computing hardware.
• Device scaling and low operating voltage are key for high-integration and energy-efficient neuromorphic systems.

Purpose of the Study:

• To fabricate and characterize sub-10 nm ferroelectric capacitors with varying HfZrO components for neuromorphic applications.
• To investigate the impact of component variation on ferroelectric properties and synaptic emulation capabilities.

Main Methods:

• Fabrication of sub-10 nm ferroelectric capacitors using HfZrO (HZO) films with varying HfO and ZrO compositions.
• Characterization of ferroelectric properties, including remnant polarization and coercive electric field (Ec).
• Evaluation of artificial synapse performance in the MNIST handwritten digit recognition task.

Main Results:

• Varying component HZO capacitors exhibited similar remnant polarization but a lower coercive electric field (Ec) compared to conventional 1:1 HZO capacitors.
• Lower Ec enables partial domain switching at lower pulse amplitudes and widths, crucial for synaptic emulation.
• Sub-10 nm ferroelectric artificial synapses achieved approximately 85.83% accuracy in the MNIST recognition task.

Conclusions:

• Varying component ferroelectric HZO capacitors offer advantages for ultra-scaled neuromorphic computing.
• These devices demonstrate potential for developing next-generation, energy-efficient artificial synapses.
• The findings pave the way for advanced neuromorphic systems utilizing ultra-scaled ferroelectric devices.