Hafnium-Based Ferroelectric Diodes for Logic-in-Memory Application.

Area of Science:

• Materials Science
• Computer Engineering
• Solid State Physics

Background:

• Traditional CMOS computing faces limitations due to the Von Neumann bottleneck, necessitating low-power in-memory logic devices.
• The development of energy-efficient computing architectures is crucial for overcoming current technological hurdles.

Purpose of the Study:

• To demonstrate ferroelectric diode devices capable of performing Boolean logic operations within memory.
• To achieve attojoule-level energy consumption for one-bit full-adder computation.
• To explore the potential of ferroelectric diodes for scalable, in-memory computing systems.

Main Methods:

• Fabrication of ferroelectric diode devices with a TiN/Hf_0.5Zr_0.5O₂/HfO₂/TiN structure.
• Implementation of 16 Boolean logic operations using single-step or multi-step cascade operations.
• Simulation of circuit schemes for logic operations and one-bit full-adder computation based on bidirectional rectification.

Main Results:

• The TiN/Hf_0.5Zr_0.5O₂/HfO₂/TiN ferroelectric diode exhibited non-destructive readout and bidirectional rectification.
• Conduction mechanism in the on-state followed Schottky emission behavior.
• Attotjoule-level energy consumption was achieved for one-bit full-adder computation.
• Logic operations and full-adder computation were successfully simulated using resistance-based input/output, eliminating the need for conversion circuits.

Conclusions:

• Ferroelectric diode devices offer a promising pathway for energy-efficient in-memory computing.
• The demonstrated logic circuits exhibit superior performance with ultralow power consumption.
• This work lays the groundwork for developing scalable and highly efficient in-memory computing systems.