Ta Phase Engineering for Defect-Controlled Reliable Switching in Ultrathin TaO_x Memristors.

Area of Science:

• Materials Science
• Solid-State Electronics
• Nanotechnology

Background:

• Ultrathin tantalum oxide (TaOx) resistive random-access memory (RRAM) and memristor devices require precise oxygen stoichiometry for reliable switching.
• The crystallographic phase of the tantalum (Ta) electrode is a critical, yet underexplored, factor influencing defect formation and device performance.

Purpose of the Study:

• To investigate how the crystallographic phase of Ta electrodes affects defect formation and switching reliability in UV-ozone (UVO)-treated TaOx devices.
• To establish electrode phase engineering as a design parameter for enhancing TaOx memristor performance for in-memory and neuromorphic computing.

Main Methods:

• Fabrication of TaOx RRAM devices using UVO oxidation on different Ta crystallographic phases (α-Ta and β-Ta).
• Comprehensive structural, chemical, and electrical characterization, including defect analysis and switching behavior evaluation.
• Analysis of filament evolution and switching stability in relation to phase-dependent defect distributions.

Main Results:

• UVO oxidation of α-Ta yields a dense, near-stoichiometric TaOx layer with a controlled substoichiometric reservoir, unlike β-Ta which forms oxygen-deficient oxides with higher trap densities.
• α-Ta-based devices demonstrate significantly reduced device-to-device and cycle-to-cycle variability compared to β-Ta counterparts.
• α-Ta-based devices exhibit improved endurance and switching stability, attributed to phase-dependent defect distributions influencing filament evolution.

Conclusions:

• The crystallographic phase of the Ta electrode critically governs defect formation and switching reliability in ultrathin TaOx memristors.
• Electrode phase engineering, specifically using α-Ta, is a key strategy for developing reliable and high-performance TaOx memristors.
• This work provides a pathway for optimizing TaOx memristors for advanced applications like in-memory and neuromorphic computing.