Research

The human brain consists of 10^11 neurons and an extensive network of 10^15 synapses. The neurons are important in the transmission of information, and the synapses act as adaptive memory elements. The high processing efficiency of the brain is closely related to the densely packed neural networks of chemical synapses which are capable of storing and processing information simultaneously. The artificial synapse is the fundamental building block for neuromorphic hardware, and emulating these synaptic characteristics are important in implementation of an efficient computational architecture. This brain-inspired computing like neuromorphic systems received lots of attention owing to low-power consumption, parallel processing, highly robust, and self-intelligent. In that sense, nanoscale devices like memristors and transistors are the front-runners with ample developments on both materials and structure for emulating advanced synaptic functions.

To realize an artificial electronic synapse that emulates the behaviour of its biological equivalent, the conductance or resistance of the electronic device (defined as synaptic weight) needs to be modulated continuously to demonstrate non-volatility. Two-terminal memristors are promising in this regard with the ability to alter conductivity based on its electronic history at an extreme low power consumption. Three-terminal transistors on the other hand offer advantages of additional gate control, thereby eliminating the need for additional training circuitry.

ReRAM devices consist of a simple capacitor-like structure, with an insulating layer in between two conducting electrodes. The origin of resistive switching in ReRAMs is usually attributed to the formation and rupture processes of nano-size conducting filaments (CFs) in the insulating layer, leading to low (SET) and high (RESET) resistive states in the ReRAM structure. The choice of electrode materials plays a key role in forming CFs in the insulating layer. With the asymmetric combination of an electrochemically active electrode (Ag or Cu) and an electrochemically inert electrode (Pt, Au, W, etc.), CFs are formed by the electrochemical dissolution and ionic migration of the electrochemically active materials, triggered by a sufficient bias voltage with proper polarity.

The realization of a fully solution based memory-resistor, deposition of an electrochemically inert electrode (bottom electrode) is the biggest challenge in achieving an all-solution-based memory, since the electrode should be electrochemically inert for ECM switching as well as thermally stable for insulating layer processing in the next growth step. We used, MIO and TiO2 were employed as the bottom electrode and the insulating layer, respectively. They were prepared by spin coating, with a polymer-assisted solution (PAS) ink. To finalize the device, an Ag solution ink was inkjet-printed as the electrochemically active electrode (top electrode). This all-solution-based ReRAM exhibits ECM-based bipolar resistive switching characteristics with high device performance and reliability, depicted by a sub-micro-second programming speed, high endurance and long retention time.