제목

RRAM-based synapse and I&F neuron device for neuromorphic pattern recognition system

내용

Hardware artificial neural network (ANN) system with high density synapse array devices can perform massive parallel computing for pattern recognition with low power consumption. To implement neuromorphic system with on-chip learning capability, we need to develop ideal synapse device with various device requirements such as scalability, multi-level cell (MLC) characteristics, low power operation, data retention, and symmetric and linear conductance change under potentiation/depression modes. In addition, we need to develop non-CMOS integrate and fire (I&F) neuron device to minimize power consumption and device area.

Various RRAM synapse devices such as filamentary switching RRAM (HfOx, TaOx) with MLC characteristics, interface switching RRAM (Pr0.7Ca0.3MnO3,TiOx) with analog memory characteristics, HfZrOx ferroelectric device and 3-terminal synapse devices using proton and oxygen migration were investigated. By optimizing forming and potentiation/depression conditions, we could improve conductance linearity and MLC characteristics of filamentary synapse device.  By controlling the reactivity of metal electrode and oxygen concentration in oxide, we can modulate the retention characteristics of interface synapse device. By separating electrodes for potentiation/depression and read mode, 3-terminal synapse exhibits better synapse performance.

Integrate and fire neuron (I&F) device were investigated using various threshold switching(TS) devices such as NbO2 based Insulator-to-Metal Transition (IMT) device, Ovonic Threshold Switching (OTS) device, and atomic-switching TS device. We found that the off-state resistance (Roff) and switching time of the TS devices determine leaky/non-leaky characteristics and activation function of neuron, respectively. Compared with conventional CMOS based I&F neuron, TS based neuron exhibits various advantages such as ultra-low power (~fJ/spike) operation, small device area and low process temperature (<400℃). 목

Based on various synapse device characteristics, we have estimated the pattern recognition accuracy of MNIST handwritten digits and CIFAR-10 dataset. We have confirmed that synapse device characteristics directly affect pattern recognition accuracy.