Abstract: Future unconventional computing systems will require computers that have high energy efficiency, low compute-memory bottleneck, are adaptive to sensory inputs, and perform analog computations on the edge. Two leading material classes to tackle these needs are atomically-thin 2D materials such as transition metal dichalcogenides (TMDs) and graphene, as well as magnetic materials based on magnet tunnel junctions (MTJs). Devices made from these materials also can make use of unique properties of the materials to compute in new ways.
We will show how the naturally ambipolar nature of TMDs can be used to construct compact circuits based on WSe2 ambipolar dual-gate transistors. Through device engineering, we measure the first WSe2 ambipolar cascaded logic circuits that do not need any additional devices to propagate the bit between transistors. By utilizing the ambipolar nature, we can also construct new circuits such as the Vt-drop circuit. We will then turn to our results on graphene-channel transistors, where the graphene conductance can be tuned using an ion-conducting Nafion layer. We will show how the devices act as artificial synapses, artificial neurons, and artificial dendrites, useful for bio-compatible neuromorphic computing. Lastly, we will show our recent results on how magnetic domain walls integrated into MTJs can act as stochastic artificial neurons, suited for computing in noisy environments on the edge. These results elucidate the wide design space for using new materials for unconventional computing.
Bio: Jean Anne C. Incorvia is an Associate Professor in Electrical and Computer Engineering at The University of Texas at Austin, where she directs the Integrated Nano Computing (INC) Lab. She received her bachelor’s in physics from UC Berkeley, and her Ph.D. in physics from Harvard University, cross-registered at MIT.