Innovation Providing New Multiple Functions in Phase-Change Materials To Achieve Cognitive Computing

Abstract

This paper describes a basic new scientific and technological approach for information and computing use. It is based on Ovonic cognitive devices that utilize an atomically engineered Ovonic chalcogenide material as the active medium. We demonstrate how such a device possesses many unique functions including an intrinsic neurosynaptic functionality that permits the processing of information in a manner analogous to that of biological neurons and synapses. Our Ovonic cognitive devices can not only accomplish conventional binary computing, but are capable of non-binary generation of information, storage, encryption, higher mathematics, modular arithmetic and factoring. Uniquely, almost all of these functions can be accomplished in a single nanosized device. These devices and systems are robust at room temperature (and above). They are non-volatile and also can include other volatile devices such as the Ovonic Threshold Switch and Ovonic multi-terminal threshold and memory devices that can replace transistors.

Memories of a visiting scientist

“…This time Stan asked me to work in his more favorable invention: the ''Ovonic memory,'' a term found even in Webster's dictionary. The Ovonic electrically erasable programmable read only memory of the 1960s, after 40 years or so of unstopped research, had transformed to the submicron devices called ''Ovonic universal memory'' (OUM) and ''Ovonic cognitive devices'' [12,13].…”

“…If their properties were optimized they could make VLSI technology much simpler and faster, could be used in an electrical or optical cognitive computer or just used for encryption of data. The most fascinating of all were the cognitive devices [12][13][14]. Their structural simplicity was amazing, but still they behaved like the biological nerves, able to make synapses.…”

Master-equation approach to understanding multistate phase-change memories and processors

“…Multilevel storage has already been demonstrated for both optical and electrical memories. 3,4 Thus, an artificial neuron might be achieved using only phase-change cells, operating in the energy-accumulating regime to mimic the operation of the neuronal body and the multilevel storage regime to mimic synaptic weighting.Several methods have been used over the years to model the process of crystallization in phase-change alloys, such as the Johnson-Mehl-Avrami-Kolmogorov ͑JMAK͒ model. Unfortunately many of the assumptions on which JMAK is based are violated in real switching events in phase-change devices.…”

“…This energy accumulation property has the potential to implement basic mathematical operations such as addition, subtraction, multiplication, and division, as well as more complex functions such as factoring, encryption, and logic. 3,4 The accumulation property, the presence of a distinct threshold, and a nonlinear ͑output͒ transition ͑between resistance states͒ mimic the basic action of a biological neuron. Furthermore, the ͑synaptic͒ weighting of inputs might be provided by another phase-change cell operating in the multilevel storage regime.…”

“…Multilevel storage has already been demonstrated for both optical and electrical memories. 3,4 Thus, an artificial neuron might be achieved using only phase-change cells, operating in the energy-accumulating regime to mimic the operation of the neuronal body and the multilevel storage regime to mimic synaptic weighting.…”