Polymer-based memristor and memristive circuits for flexible electronic system

Abstract

Flexible electronics such as smart band, smart watch, and curved display have been extensively investigated as the next generation electronics due to its innovative design, user-friendly interfaces, and convenience portability. However, as the flexible electronic systems have a limited battery supply and long standby period, it is essential for building up low-power flexible electronic system. This thesis focuses on developing a nano electronic device which enables a novel low-power computing architecture. For that purpose, the flexible memristor array based on poly(1,3,5-trimethyle-1,3,5-trivinyl cyclotrisiloxane) (pV3D3) deposited via initiated chemical vapor deposition (iCVD) have been developed. The developed pV3D3-memristor showed outstanding electrical characteristics and good uniformity in terms of device-to-device distribution compared to solution-processed polymer-memristors. Also, the interfacial engineering technique using multilayer graphene insertion layer has been devised for reduction of power consumption and improvement of resistive switching uniformity.

Moreover, we implemented the nonvolatile logic-in-memory circuit with static power consumption of 0 W using the pV3D3-memristor array, and realized a half adder as well as the basic Boolean logic gates. Also, to resolve an inherent sneak current problem that causes cell-to-cell interference in crossbar array, we developed an amorphous In-Zn-Sn-O (a-IZTO) semiconductor-based back-to-back Schottky diode type selector device and integrated with the pV3D3-memristor. Thus, we successfully implemented a single-instruction multiple-data (SIMD), which is the foundation of parallel computing, using the flexible 1S-1M array.

Finally, we revealed the origin of the digital and analog switching of the filament type memristor using the quantized conductance and geometry of Cu filament using TEM. Thereby, we implemented the analog switching in the pV3D3-memristor and emulated the synaptic behaviors essential for learning process such as synaptic plasticity and spike-time dependent plasticity. Moreover, to verify the feasibility for pattern recognition application, we constructed an artificial neural network based on pV3D3-mermistor and evaluated the face classification using device-to-system level framework.