Thermal Management of Nanoscale Memristors

Position – Post-Doctoral Fellow

Host – Electronics Manufacturing and Reliability Laboratory 

PI – Dr. Samuel Graham 

Duration – July 2018 – Present

Motivation and Objectives

The primary unit of currently used digital computers are field-effect transistors that perform Boolean functions. They are excellent at performing complex arithmetic and logic calculations, but lag far behind the human brain in key areas such as adaptivity, generalization, and pattern recognition. Adaptive oxides of Niobium, Tantalum, Vanadium  are key candidates for neuromorphic behavior including memory effects or memristive abilities. It is essential therefore to understand how these materials exhibit a diode-like behavior, wherein they offer high electrical resistance for a range of voltage and then very low resistance after the voltage exceeds a threshold. This 2-terminal structure is fairly simple than the conventional 3-way semiconductors, however the exact mechanisms why this switching takes place in these oxides is not known.  Understanding, modeling and experimentally validating these mechanisms is the key objective of this work, while making an energy efficient memristor network becomes another task.

Sample Results 

While the overall goal is still in the uncharted territory, the following animation shows a sample geometry modeled in COMSOL Multiphysics, where joule heating as a result of current flow through an adaptive oxide is shown. Increase in temperature due to joule heating increases the electrical conductivity in the oxides, which in turn increases the current flow, which then increases the temperature. This positive feedback loop is of primary interest to several research thrusts in this area.

Temperature contours of a single memristor device. Voltage across it increases linearly with time and then decreases linearly.

This is still a developing story and more results will follow.