Low-Power Microwave Relaxation Oscillators Based on Phase-Change Oxides for Neuromorphic Computing

Abstract

Neuro-inspired computing architectures are one of the leading candidates to solve complex, large-scale associative learning problems. The two key building blocks for neuromorphic computing are the synapse and the neuron, which form the distributed computing and memory units. Solid state implementations of these units remain an active area of research. Specifically, voltage or current controlled oscillators are considered a minimal representation of neurons for hardware implementations. Such oscillators should demonstrate synchronization and coupling dynamics for demonstrating collective learning behavior, besides the desirable individual characteristics such as scaling, power, and performance. To this end, we propose the use of nanoscale, epitaxial heterostructures of phase change oxides and oxides with metallic conductivity as a fundamental unit of an ultralow power, tunable electrical oscillator capable of operating in the microwave regime. Our simulations show that optimized heterostructure design with low thermal boundary resistance can result in operation frequency of up to 3 GHz and power consumption as low as 15 fJ/cycle with rich coupling dynamics between the oscillators.

Emerging Artificial Neuron Devices for Probabilistic Computing

“…Further research indicated that the applied voltage can regulate the output frequency (Zhang et al, 2020), as depicted in Figure 10D. Additionally, a microwave oscillator circuit was proposed (Zhao and Ravichandran, 2019) to generate output oscillation frequencies as high as 3 GHz with energy consumption as low as 15 fJ/spike. Furthermore, the output voltage frequency can be adjusted based on the external pressure by coupling the device with an afferent sensor, such as a piezoelectric device (Figure 10E).…”

Origin of Current‐Controlled Negative Differential Resistance Modes and the Emergence of Composite Characteristics with High Complexity

“…Current-controlled negative differential resistance (NDR) is of increasing interest for braininspired computing based on a number of promising applications, including trigger comparators 1 , self-sustained and chaotic oscillators [2][3][4][5][6][7] , threshold logic devices 8,9 and the emulation of biological neuronal dynamics 10,11 . Since the functionality of such devices is determined by the specific form of their current-voltage characteristics, it is important to understand the physical basis of switching and how it is affected by device structure and operating conditions.…”

Progress and perspective on polymer templating of multifunctional oxide nanostructures

“…7 For the latter, memristive oxides (e.g., NbO x or VO 2 ), which can switch between various resistance states, are attracting much attention. [8][9][10][11][12][13][14][15][16][17] Nanocomposite materials using two different oxides can be desirable to take advantage of the complementary properties of both compounds (e.g., piezo-pyroelectric composites 18 ) or in order to achieve large interfacial areas and enhance the coupling of their properties or their connectivities. Despite their promise, fully inorganic nanocomposites are not broadly investigated.…”