Tuning Oxygen Vacancy Dynamics in Al₂O₃ Memristors via Mg and Zn Doping
ORAL
Abstract
Memristors, which possess both storage and processing capabilities, are crucial components for neuromorphic computing and resistive random-access memory (RRAM). A promising class of memristors relies on atomic-scale control of oxygen vacancies (VO) in metal-oxide thin films to modulate resistive switching behavior.
Density functional theory (DFT) simulations using hybrid functionals reveal that doping pristine Al₂O₃ with Mg lowers the Fermi energy, thereby enhancing the high-resistance state (HRS) and reducing the oxygen vacancy formation energy to promote conductive filament (CF) formation. Experimental results using in vacuo atomic layer deposition confirm these predictions by demonstrating exponential increases in HRS and on/off ratios (10–10⁴) with the insertion of an MgO layer [1,2].
Building on these insights, we investigated alternative dopants such as Ca, Sr, Be (from the same group as Mg), and Zn (with comparable ionic radii), as well as d-block elements like Cd, Cu, and Ni. Among these, Zn emerges as a particularly promising candidate by showing similar performance as Mg.
These findings advance the understanding and atomic-level design of high-performance, energy-efficient memristors for next-generation neuromorphic systems.
[1] Marshall et al., ACS Applied Materials & Interfaces 17, 49930 (2025).
[2] Goul et al., Communications Physics 5, 260 (2022).
Density functional theory (DFT) simulations using hybrid functionals reveal that doping pristine Al₂O₃ with Mg lowers the Fermi energy, thereby enhancing the high-resistance state (HRS) and reducing the oxygen vacancy formation energy to promote conductive filament (CF) formation. Experimental results using in vacuo atomic layer deposition confirm these predictions by demonstrating exponential increases in HRS and on/off ratios (10–10⁴) with the insertion of an MgO layer [1,2].
Building on these insights, we investigated alternative dopants such as Ca, Sr, Be (from the same group as Mg), and Zn (with comparable ionic radii), as well as d-block elements like Cd, Cu, and Ni. Among these, Zn emerges as a particularly promising candidate by showing similar performance as Mg.
These findings advance the understanding and atomic-level design of high-performance, energy-efficient memristors for next-generation neuromorphic systems.
[1] Marshall et al., ACS Applied Materials & Interfaces 17, 49930 (2025).
[2] Goul et al., Communications Physics 5, 260 (2022).
*This work was supported by the National Science Foundation grant DMR-2425549.
–
Presenters
-
YASHASSRI RAJIND BALASOORIYA GALGAMU ARACHCHILLAGE
- University of Kansas