Abstract
The use of BaTiO3 (BTO) ferroelectric thin films in flexible ferroelectric memory offers a promising pathway for next-generation nonvolatile memory applications, given BTO’s excellent ferroelectric properties, stability, high dielectric constant, and strong fatigue resistance. However, the fabrication of BTO on flexible substrates presents a significant technical challenge. In this study, we achieved high-quality, single-crystalline (111)-oriented BTO films on mica substrates through the design of buffer layers. The BTO films exhibit strong polarization properties (remnant polarization, 2Pr ∼15.63 μC/cm2, and saturation polarization, 2Ps ∼36.61 μC/cm2), and the flexible BTO devices maintained exceptional stability under bending radii of 3.5 and 6 mm. After 107 bipolar switching cycles, polarization showed only minor changes, with a retention time exceeding 104 s. We further explored the application of flexible BTO ferroelectric memory in neuromorphic computing. The flexible BTO-based memory demonstrated adjustable synaptic behavior, effectively modulating EPSC (excitatory postsynaptic current) responses through pulse amplitude and width to simulate short-term memory. PPF (paired pulse facilitation) and LTP (long-term potentiation) behaviors verified its synaptic weight modulation capabilities, achieving 91.6% accuracy in neural network-based handwritten digit recognition after 103 training cycles. These findings underscore the potential of flexible BTO ferroelectric memory for memory devices and neuromorphic computing, offering promising applications for wearable AI systems.
| Original language | English |
|---|---|
| Pages (from-to) | 18571-18581 |
| Number of pages | 11 |
| Journal | ACS Applied Materials and Interfaces |
| Volume | 17 |
| Issue number | 12 |
| DOIs | |
| State | Published - 26 Mar 2025 |
Keywords
- AI systems
- BTO film
- ferroelectric memory
- flexible
- neuromorphic computing
- synaptic behavior
Fingerprint
Dive into the research topics of 'Flexible BaTiO3 Ferroelectric Nonvolatile Memory for Neuromorphic Computation'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver