Sub-10nm Ultra-thin ZnO Channel FET with Record-High 561 µA/µm I_ON at V_DS 1V, High µ-84 cm²/V-s and1T-1RRAM Memory Cell Demonstration Memory Implications for Energy-Efficient Deep-Learning Computing

Abstract

For the first time, we investigated ultra-short-channel ZnO thin-film FETs with L <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ch</inf> = 8 nm with extremely scaled channel thickness t <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ZnO</inf> of 3nm, the device exhibits ultra-low sub-pA/µm off leakage (1.2 pA/µm), high electron mobility (µ <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">eff</inf> = 84 cm2/V•s) with record peak transconductance (Gm,) of 254 μS/μm at V <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">DS</inf> = 1 V wrt. reported oxide-based transistors, to date, leading to high on-state current (I <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ON</inf> ) of 561 μA/μm. We demonstrated the integration of a ZnO access transistor with Al <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> O <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</inf> RRAM to enable a 1T-1R memory cell, suitable for BEOL-embedded memory. We evaluate the system-level benefits of a hardware accelerator for deep learning to employ FET-RRAM as working memory—up to 10X energy-efficiency benefits can be achieved over current baseline configurations.