A Scalable Distributed Method for Quantum-Scale Device Simulation
A Scalable Distributed Method for Quantum-Scale Device Simulation
S. Cauley, J. Jain, C.-K. Koh and V. Balakrishnan
Abstract: We present an algorithm for the fast and accurate simulation of
nano-scale devices not attainable using current techniques. The idea
underlying the algorithm is a novel divide-and-conquer method based on
the non-equilibrium Green's function formalism. This formalism has
provided a unifying conceptual framework for the analysis of quantum
transport in nanodevices and the computations therein are of
significant interest across many areas of research. We offer two
applications, the atomistic level simulation of silicon nanowires and
the two-dimensional simulation of nanotransistors, which highlight the
benefits of the divide-and-conquer framework. The inherently parallel
algorithm presented here allows for computing resources to be flexibly
allocated toward either solving problems of larger sizes in comparable
time, or speeding up the solution of a problem for a given size. Our
algorithm facilitates the solution of problems orders of magnitude
larger, and in most cases was able to achieve substantial speed-up, as
compared to the current state of the art algorithm. We conclude that
the method presented here allows for large-scale simulation problems
that can now be realized without the use of special purpose hardware
or approximation methods.
Download PDF
Bibtex entry