Theoretical Models of Quantum Conductance on Graphical Processing Units

ABSTRACT: Due to the costs associated with their fabrication, theoretical prediction of conductance in molecular and nano-electronic systems is a cost-effective tool in the design of new devices. Unfortunately, the techniques currently used to treating conductance are computationally expensive, and modeling real systems accurately has been a challenge. We propose to tackle this problem by developing new and more effective algorithms and exploring new computer architectures. Research at UC-Davis has yielded a technique (Tunneling Current, TC, theory) for efficient calculation of conductance in complex molecules and nano-structures. In addition, PetaChem, LLC now maintains quantum chemistry code (TeraChem) designed for graphical processing units (GPUs) that calculates electronic structure in systems as large as 2000 atoms using DFT. Our Phase I effort will focus on combining these two advances to improve the reliability of theoretical modeling for nano-electronic devices. Specifically, TC theory will be tested for its accuracy in modeling conductance in metal-molecule-metal junctions, and results will be compared to those from RT-TDDFT. Algorithms suitable for GPU implementation of TC theory will then be derived. Our Phase II product will be computer software that allows both TC and RT-TDDFT calculations and executes on GPU platforms. BENEFIT: The product of the proposed STTR effort, after Phase II, is a computer software module that would allow for the efficient and accurate prediction of the linear and non-linear optical response of materials. The software, which will be designed to run on graphical processing units, will permit efficient, thousand-atom simulations using the high-level real-time TDDFT method and periodic boundary conditions. These simulations will be 2-3 times faster and 10 times bigger than those performed by current software on standard CPU platforms. The Phase I proof-of-principle demonstration will consist of a test of full-response function, real-time (RT) TDDFT for its applicability to Air Force problems and the derivation of the mathematical algorithms for RT-TDDFT with periodic boundary conditions. This new simulation software will enable modeling of the complex interaction of light and matter in photovoltaic materials and non-linear optical devices, as well the UV/Vis absorption spectra of chemical and biological agents, explosives for better detection techniques.