CMOS-Compatible LiNbO₃ Domain-Wall Entropy Engine for Ultrafast True Random Number Generation

Area of Science:

• Materials Science
• Condensed Matter Physics
• Semiconductor Devices

Background:

• Ferroelectric materials exhibit random domain nucleation and growth, forming high-entropy systems.
• Conventional true random number generation (TRNG) using electrical methods faces limitations due to low noise amplitudes.
• Existing TRNG methods struggle with signal amplification and operational robustness.

Purpose of the Study:

• To develop a novel TRNG platform utilizing stochastic conductive domain walls in ferroelectric thin films.
• To overcome the noise amplitude limitations of conventional electrical TRNG methods.
• To establish a robust and scalable entropy source for secure random number generation.

Main Methods:

• Fabrication of single-crystal LiNbO3 thin films bonded to SiO2/Si wafers.
• Utilizing stochastic formation and switching of conductive domain walls as the entropy source.
• Characterization of noise amplitude, cycling endurance, and switching speeds of the domain wall device.

Main Results:

• Achieved high noise amplitude (>500 nA) and exceptional cycling endurance (>10^10).
• Demonstrated ultrafast, self-reinitialized stochastic domain switching operations (<10 ns).
• Generated random bitstreams validated by NIST tests, showing resistance to machine learning attacks.

Conclusions:

• Ferroelectric conductive domain walls provide a groundbreaking platform for CMOS-compatible entropy source extraction.
• This approach effectively addresses challenges in amplifying entropy signals with operational robustness.
• The developed TRNG method offers a promising solution for ultrafast and secure random number generation.