Monolithic Oxidation Enables Ultrathin Vertically Graded Tantalum Oxide for Low‐Voltage, Low‐Variability Memristive Switching
ABSTRACT Artificial intelligence (AI) workloads are increasingly bottlenecked by the energy and latency from data movement between memory and logic, necessitating intermediate hardware solutions such as compute‐in‐memory (CIM). Metal oxide‐based synaptic memristors are promising for CIM because their arrays perform vector‐matrix multiplication. These devices must operate at core‐supply voltages below 1 V with low‐variability analog updates compatible with standard complementary metal‐oxide‐semiconductor platforms. However, simultaneously achieving low voltage and low variability remains challenging; thicker oxides can improve uniformity but demand higher voltages, whereas ultrathin films reduce voltage at the cost of increased variability. Here, we report a monolithic, room‐temperature ultraviolet‐ozone (UVO) oxidation process that enables an ultrathin (∼3.5 nm), stratified Ta 2 O 5 /TaO x layer with a vertical stoichiometry gradient. This self‐limiting, deposition‐free growth approach yielded balanced low‐voltage operation with controlled variability, yielding sub‐1 V SET and RESET voltages ( V SET ≈ −0.4 V and V RESET ≈ 0.7 V) with low cycle‐to‐cycle variability (σ SET ≈ 0.05 V and σ RESET ≈ 0.01 V). Gradual, repeatable conductance updates demonstrated stable analog programmability for inference. Synchrotron variable‐energy X‐ray photoelectron spectroscopy confirmed growth saturation and vertically graded ultrathin oxide. The demonstrated UVO monolithic oxidation and resulting ultrathin graded TaO x structure provide a scalable route to uniform low‐voltage memristive switching.
Authors
- Aswin kumar Anbalagan (ORCID: https://orcid.org/0000-0001-5511-2083)
- Nikhil Tiwale (ORCID: https://orcid.org/0000-0001-8229-7108)
- Houk Jang (ORCID: https://orcid.org/0000-0002-7236-1134)
- Chang‐Yong Nam (ORCID: https://orcid.org/0000-0002-9093-4063)
- Mingzhao Liu (ORCID: https://orcid.org/0000-0002-0999-5214)
- Xiao Tong (ORCID: https://orcid.org/0000-0002-5567-9677)
- Conan Weiland (ORCID: https://orcid.org/0000-0001-6808-1941)
- Won‐Il Lee (ORCID: https://orcid.org/0000-0003-1043-9933)
- Seunghoon Yang (ORCID: https://orcid.org/0000-0002-2447-4062)
- Dong Hyun Lee
- Soumyajit Mandal
- Kim Kisslinger
Institutions
- National Institute of Standards and Technology (US)
- Brookhaven National Laboratory (US)
- Material Measurement Laboratory (US)
Publication Details
- Journal
- Advanced Functional Materials
- Published
- 2026-07-23
- DOI
- https://doi.org/10.1002/adfm.77144
- Primary Topic
- Advanced Memory and Neural Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.00