Low-cost, CMOS compatible, Ta2O5-based hemi-memristor for neuromorphic circuits

Abstract

In the past, tantalum oxide devices have been used to create non-volatile digital memories, whilst neglecting the analogue memristive characteristics of such devices. In this Letter, it is shown that these devices can provide a low-cost, low-power solution for hemi-memristive devices, when used in their pre-formed, memristive region, whilst being fully CMOS compatible. Furthermore, measurements are presented from the devices that have been fabricated and it is shown that these devices do not require electroforming. Electroforming circuitry takes up valuable chip space at the transistor layers and significantly increases fabrication cost, since voltages as high as 12 V are required, which in turn requires extra masks to form high voltage devices and distribution circuits.

Sound Source Localization through 8 MEMS Microphones Array Using a Sand-Scorpion-Inspired Spiking Neural Network

“…Additionally, the results indicate that, by varying different time parameters of the neural network, the performance of the SNN can be improved, and most probably reach the performance reported for the analog neural network, where the error is in the low single-digit range and the standard deviation at 4.5° (van Schaik and Shamma, 2003, 2004) or even the ASU units with CMOS VLSI design, which is reported to be around 1°–2° (Julian et al, 2006; Goodman and Brette, 2008; Chacon-Rodriguez et al, 2009). As the neural network is very simple it is expected that the performance parameters are advantageous, which we want to investigate and prove an implementation in custom hardware, using a CPLD/FPGA or even new memristive devices (Kyriakides et al, 2012) is feasible.…”

Solution-processed, low voltage tantalum-based memristive switches

“…Anodization is a self-limiting and self-healing process that gives pinhole-free, homogenous oxide layers grown in an ambient atmosphere at room temperature using low cost set-ups and environmentally friendly chemicals. Tantalum/tantalum oxide (Ta/TaOx) memristive switches have been reported to show superior performance compared to other structures in terms of volatility [6], low voltage [7] and switching times [8]. They have been investigated by researchers to be used in ReRAM applications and more recently in neuromorphic computing since they can be used as neural synapses [9,10].…”

Memristor-Based Loop Filter Design for Phase Locked Loop

“…A loop filter eliminates the high frequency components from the phase comparator so that only the direct current component is provided to the VCO. A memristor is considered a nanoscale device, so it is useful for more applications such as nonvolatile memory applications, low power and remote sensing applications, cross bar latches as transistor replacements, and analog computation and circuit applications [4]. Memristors' ability to maintain a state without requiring external biasing can significantly reduce overall power consumption, while the deep-nanoscale physical dimensions of the device (minimum reported: 5 × 5 nm) are ideal for its implementation in the field of VLSI (very large-scale integration) [5] and can thus provide a much-needed extension to Gordon Moore's law.…”