SBIR-STTR Award

The behavior of the Chalcogenide based ion-conducting memristor lends itself for use as an element in a simple neuromorphic computing circuit. The reaction of a circuit to an external stimulus may be the result of its ability to learn from previous similar, yet unrelated exposures to environmental stimulus. A highly specialized variation of the memristor, previously developed by the Advanced Memory and Reconfigurable Logic Group at Boise State University, led by Dr. Kris Campbell, PhD. will be leveraged as the key functional component in a family of neuromorphic circuits. The electrical memristor device has demonstrated predictable discrete states, hysteresis, and time dependent memory, each of which may be leveraged for functional decision behaviors. In essence, the device remembers past stimulus, has a current state that is a direct function of both the past stimulus and the current stimulus, and eventually forgets, depending on the strength of the programming state and the passage of time. These unique device characterisics enable a new classification of intellegent computing architectures that may have the potential to demonstrate rudimentary adaptive learning behaviors similar to those found in nature.

Benefit:
The research supported will enable the design and prototyping of functional memristor based neuromorphic circuits which may potentually revolutionize the electronics industry by introducing new methodologies facilitating neuromorphic computing. These memristor based circuits will be demonstrated and integrated within existing technology frameworks, establishing the feasibility of manufacture within current technology nodes. The IP cores and demonstration vehicles developed during the course of this work are directly followed by a commercialization strategy to quickly expand the fledgling memristor efforts in industry. The result is intended to enable ground breaking applications and exploration in neuromorphic computing and reconfigurable electronics.

Keywords:
Neuromorphic Computing, Memristor, Electronic Intelligence, Adaptive Learning Behavior, Chalcogenide, Ion-Conducting, Reconfigurable, Memory

The chalcogenide based ion-conducting memristor has been shown to be an effective element as the core of a simple neuromorphic computing circuit. The response of the computing circuit is the result of weighted external stimuli, the current state of the device, and the history of exposure by the device to the stimuli. A highly specialized version of the device, as developed by the Advanced Memory and Reconfigurable Logic Group at Boise State University led by Dr. Kris Campbell, PhD. will be designed, fabricated, and integrated to form the core of a “plug and play” soft processor microchip. Leveraging the unique properties of the memristor device, the soft processor will exploit the nature-like learning capabilities of a memristor based synthetic synapse to form an adaptive, responsive, rudimentary learning system on a chip. The synapses, fabricated to form an addressable neural fabric will be fully integrated into conventional digital support architectures, including standard buffers, registers, a communications BUS, and input/output ports. Device design and fabrication will be developed using industry standard tools and well established technology nodes, enabling quick and effective integration into the mainstream electronics community, providing opportunities for unprecedented advances in intelligent computing architectures, adaptive analysis, and responsive control systems.

Benefit:
The primary technical objective of the Phase II work is the design and fabrication of a fully functional Plug-and-Play “soft processor” chip based on the chalcogenide ion-conducting thin film memristor tested in the Phase I effort. The commercial target is straightforward: develop a “soft processor” designed using chalcogenide based ion-conducting thin film memristors, operating as a hardware based, high speed neural network, capable of fuzzy logic decision and classification of input signals. The design and fabrication is intended to be fully industry compatible, utilizing industry recognized tools, technology nodes, and methodologies. The final component is intended to be easily integrated into commercial circuit designs. The objective is to provide the electronics industry with a self-contained functional memristor based building block that can be applied by any designer with reasonable skills and design tools. The expectation is that the building block chip enabled by this Phase II work will enable a large variety of engineers and scientists to develop applications currently unimagined with existing technology limitations.

Keywords:
Neuromorphic Computing, Memristor, Ion Conducting, Adaptive Analysis, Chalcogenide, Soft Processor, Reconfigurable, Synthetic Neuron