Kibong Moon

Neuromorphic Computing Using Non-Volatile Memory

Abstract Dense crossbar arrays of non-volatile memory (NVM) devices represent one possible path for implementing massively-parallel and highly energy-efficient neuromorphic computing systems. We first review recent advances in the application of NVM devices to three computing paradigms: spiking neural networks (SNNs), deep neural networks (DNNs), and ‘Memcomputing’. In SNNs, NVM synaptic connections are updated by a local learning rule such as spike-timing-dependent-plasticity, a computational approach directly inspired by biology. For DNNs, NVM arrays can represent matrices of synaptic weights, implementing the matrix–vector multiplication needed for algorithms such as backpropagation in an analog yet massively-parallel fashion. This approach could provide significant improvements in power and speed compared to GPU-based DNN training, for applications of commercial significance. We then survey recent research in which different types of NVM devices – including phase change memory, conductive-bridging RAM, filamentary and non-filamentary RRAM, and other NVMs – have been proposed, either as a synapse or as a neuron, for use within a neuromorphic computing application. The relevant virtues and limitations of these devices are assessed, in terms of properties such as conductance dynamic range, (non)linearity and (a)symmetry of conductance response, retention, endurance, required switching power, and device variability. Graphical Abstract

Improved Synaptic Behavior Under Identical Pulses Using AlO x /HfO 2 Bilayer RRAM Array for Neuromorphic Systems

We analyze the response of identical pulses on a filamentary resistive memory (RRAM) to implement the synapse function in neuromorphic systems. Our findings show that the multilevel states of conductance are achieved by varying the measurement conditions related to the formation and rupture of a conductive filament. Furthermore, abrupt set switching behavior in the RRAM leads to an unchanged conductance state, leading to degradation in the accuracy of pattern recognition. Thus, we demonstrate a linear potentiation (or depression) behavior of conductance under identical pulses using the effect of barrier layer on the switching, which was realized by fabricating an RRAM on top of an Al electrode. As a result, when the range of the conductance is symmetrically controlled at both polarities, a significantly improved accuracy is achieved for pattern recognition using a neural network with a multilayer perceptron.

Structurally Engineered Stackable and Scalable 3D Titanium‐Oxide Switching Devices for High‐Density Nanoscale Memory

A 3D high-density switching device is realized utilizing titanium oxide, which is the most optimum material, but which is not practically demonstrated yet. The 1S1R (one ReRAM with the developed switching device) exhibits memory characteristics with a significantly suppressed sneak current, which can be used to realize high-density ReRAM applications.

TiO x -Based RRAM Synapse With 64-Levels of Conductance and Symmetric Conductance Change by Adopting a Hybrid Pulse Scheme for Neuromorphic Computing

We propose TiOx-based resistive switching device for neuromorphic synapse applications. This device is capable of 64-levels conductance states because of their optimized interface between the metal electrode and the TiOx film. To compensate the change in switching power with increasing pulse number, we propose the use of fixed voltage and current pulses in potentiation and depression conditions, respectively. By adopting a hybrid pulse scheme, the symmetry of conductance change under both potentiation and depression conditions is shown to be significantly improved. Both the improved conductance levels and the symmetry of conductance change are directly related with enhanced pattern recognition accuracy, which is confirmed by a neural network simulation.

Effects of RESET Current Overshoot and Resistance State on Reliability of RRAM

Current overshoot has severe effects on the reliability of resistive random access memory (RRAM). It is well known that the current overshoot during the SET process is caused by parasitic capacitance. In this letter, we observed a different type of current overshoot during the RESET process. The RESET current overshoot was confirmed to have severe effects on the endurance of RRAM. We also demonstrated the relation between the current overshoot and the intrinsic capacitive elements of each state of RRAM. Finally, an optimized pulse shape was proposed to minimize the current overshoot and was experimentally verified to significantly improve the variability and endurance in a typical RRAM device with a W/Zr/HfO2/TiN structure.

Highly Reliable Resistive Switching Without an Initial Forming Operation by Defect Engineering

The effects of stack and defect engineering of metal-oxide layers on resistive switching uniformity were investigated to obtain resistive random access memory (ReRAM) with excellent switching reliability. Uniform switching, parameters, such as set voltage (Vset), reset voltage (Vreset), low-resistance state, high-resistance state, and retention characteristics, were significantly improved by stack and defect engineering. Furthermore, the initial forming operation, which is a nuisance, was removed to realize cross-point ReRAM.

Optimized Programming Scheme Enabling Linear Potentiation in Filamentary HfO 2 RRAM Synapse for Neuromorphic Systems

In this brief, we demonstrate the multilevel cell (MLC) characteristics of an HfO2-based resistive memory (RRAM) array as a synaptic element for neuromorphic systems. We utilize various programming schemes to linearly change the resistance state with either set voltage/pulse ramping or gate voltage ramping. Our results reveal that the MLC relates to the size of the conductive filament involved in the movement of oxygen vacancies with respect to applying pulses. Thus, by optimizing the pulse for a set condition, such as an identical pulse, we achieve linearly increased MLC behavior, thereby enabling a high accuracy for pattern recognition in neuromorphic systems.

Oxide based nanoscale analog synapse device for neural signal recognition system

We report oxide based analog synpase for neuromorphic system. By optimizing redox reaction at the metal/oxide interface, we can obtain stable analog synapse characteristics and wafer scale switching uniformity. We have confirmed the feasibility of neuromorphic hardware system with oxide synapse array device which recognizes the electroencephalogram (EEG) signal and rats neural signal.

Steep Slope Field-Effect Transistors With Ag/TiO 2 -Based Threshold Switching Device

In this letter, we demonstrate a steep slope field-effect transistor (FET) using a threshold switching (TS) device. The Ag/TiO2-based TS device reported in our previous work was implemented in series with the drain region of a transistor. Since the TS device has an abrupt transition between the OFF- and ON-states and vice versa, the transistor has a 5-mV/decade subthreshold slope and a high ON/OFF-current ratio (ION/IOFF) of >107 with a low drain voltage (0.3 V). Furthermore, the threshold voltage (Vth,FET) of the transistor can be tuned by controlling the thickness of the TS device.

High density neuromorphic system with Mo/Pr0.7Ca0.3MnO3 synapse and NbO2 IMT oscillator neuron

We report novel nanoscale synapse and neuron devices for ultra-high density neuromorphic system. By adopting a Mo electrode, the redox reaction at Mo/Pr0.7Ca0.3MnO3 (PCMO) interface was controlled which in turn significantly improve synapse characteristics such as switching uniformity, disturbance, retention and multi-level data storage under identical pulse condition. Furthermore, The NbO2 based Insulator-Metal Transition (IMT) oscillator was developed for neuron application. Finally, we have experimentally confirmed the realization of pattern recognition with high accuracy using the 11k-bit Mo/PCMO synapse array and NbO2 oscillator neuron.