Bending Resistant Multibit Memristor for Flexible Precision Inference Engine Application

This work reports 2-bits/cell hafnium oxide-based stacked resistive random access memory devices fabricated on flexible polyimide substrates for neuromorphic applications considering the high thermal budget. The ratio of low-resistance state current (<inline-formula> <tex-math notation="LaTeX">$\textit{I}_{\biosc{on}})$</tex-math> </inline-formula> to high-resistance state current (<inline-formula> <tex-math notation="LaTeX">$\textit{I}_{\biosc{off}})$</tex-math> </inline-formula> or <inline-formula> <tex-math notation="LaTeX">$\textit{I}_{\biosc{on}}/\textit{I}_{\biosc{off}}$</tex-math> </inline-formula> for the fabricated devices was above 1.4 <inline-formula> <tex-math notation="LaTeX">$\times$</tex-math> </inline-formula> 10<inline-formula> <tex-math notation="LaTeX">$^{\text{3}}$</tex-math> </inline-formula> with a low device-to-device variation at 100 <inline-formula> <tex-math notation="LaTeX">$\bm{\mu}$</tex-math> </inline-formula>A current compliance. The mechanical stability over 10<inline-formula> <tex-math notation="LaTeX">$^{\text{4}}$</tex-math> </inline-formula> bending cycles at a 5 mm bending radius and endurance over 10<inline-formula> <tex-math notation="LaTeX">$^{\text{6}}$</tex-math> </inline-formula> WRITE cycles makes these devices suitable for online neural network training. The data retention capability over 10<inline-formula> <tex-math notation="LaTeX">$^{\text{4}}$</tex-math> </inline-formula>s at 125 <inline-formula> <tex-math notation="LaTeX">$^{\bm{\circ}}$</tex-math> </inline-formula>C also infuses these devices&#x2019; long-term inference capability. Furthermore, the performance of the devices has been verified for neuromorphic applications by system-level simulations with experimentally calibrated data. The system-level simulation reveals only a 2% loss in inference accuracy over ten years from the baseline.