NANOTECHNOLOGY
Current Research - Quantum-Dot Arrays: Simulation Results Toward Analog Computing
Summary
Arrays of quantum dots (QD) produce bi-stable and multi-stable robust behavior, which can be harnessed for unconventional, yet powerful computational concepts. We are pursuing a novel approach to signal pattern analysis using an array of QD. Our methodology combines an ultrafast neuromorphic-learning algorithm with photon-assisted tunneling in the QD array. The latter enables emulation of the plasticity of neural synapses. Simulation results illustrate the feasibility of the approach. We also compute the configuration and electronic structure of the actual 2-nm metal clusters to serve as the quantum dots by first-principle computational simulation, including the characterization of the charging characteristics of these nanoclusters, the innovative device architectures, and microscopic transport models for the device operation.
Recent developments in nanoscale science and technology have opened exciting opportunities for revolutionary advances in nanoscale computing, communication, detection, and sensing. Fully exploiting this emerging potential requires a deep understanding of the complex dynamics and properties of small arrays of quantum structures, including quantum dots (QD), ultra small Josephson junctions, QD lasers, and others. Such arrays are known to produce robust bi-stable and multi-stable behavior, which can be harnessed for unconventional, yet powerful computing. This collaborative research is focused on the basic science and engineering issues of complex information processing by QD arrays. In particular, we are interested in demonstrating a capability for pattern classification using neuromorphic algorithms.
We are exploring an opportunity for implementing neuromorphic algorithms for pattern recognition in QD arrays of sufficiently close so that transport through the array occurs via single-electron tunneling. Previous work on this topic is discussed, based on the formal analogy between the Lyapunov function of a neural network and the electrostatic free energy of a quantum-dot array. While attractive, it is practically difficult to leverage this analogy for pattern recognition because of one’s inability to modify the capacitance of the QD array after fabrication of the device. One possibility around this obstacle is to interface a gate electrode to each (or to many) QDs in the array. This also has tremendous technical difficulties for QD of a size appropriate for room-temperature operation (~ 1 nm). An alternative approach is to modulate the current-voltage relations of the QD array by interaction with an oscillatory applied field. We have presented simulation results that demonstrate that various parameters of the field (i.e., frequencies and amplitudes within realistic ranges) may be used to modulate the current, and therefore utilized as control parameters for the neuromorphic pattern recognition algorithms. In this manner, rather than insisting on using external controls local in space (which are not practical), we choose external controls that are localized in a reciprocal space (e.g., frequency space).
Full Article
Click below for PDF file
Authors:
J. C. Wells, J. Barhen, Y. Braiman, V. Protopopescu, D. J. Dean*, T. Papenbrock*
Center for Engineering Science Advanced Research
Computer Science and Mathematics Division
*Physics Division
Oak Ridge National Laboratory
W. Andreoni and A. Curioni
IBM Zurich Research Laboratory
Rueschlikon, Switzerland
[