Scalable First Principles Methods for Electronic Transport
Developing scalable methods (able to fully utilize petascale systems) enables predictive simulations of entire device structures from first principles. These methods are revolutionizing the design of molecular and nanoscale sensors and electronic devices.
Oak Ridge National Laboratory (ORNL) has developed a scalable first principles approach for quantum transport where it has developed a multi-level parallel sparse iterative method for solving the non-equilibrium Green’s function and implemented and optimized a first principles electronic structure method. Both methods exhibit high scalability and serial performance with both increasing system size and processor count.
To test these methods, ORNL is examining a system involving the inclusion of light absorbing antenna chromophores through a covalent linkage combined with the extended π electrons of a carbon nanotube. This system can constitute an ideal nano-assembly for generating singlet excited energy and its conversion to chemical energy. This type of system is of significant importance for solar energy applications.
Simulations on systems sizes up to 6128 atoms.
The test system represents a true computational challenge for a first principles approach because the system ranges in size from 1532 atoms to over 6000 atoms. To address the electronic structure ORNL implemented the Pseudopotential Algorithm for Real-Space Electronic Calculations (PARSEC) (http://www.ices.utexas.edu/parsec/index.html) on the Cray XT4 at the National Center for Computational Sciences located at Oak Ridge National Laboratory.
These methods would impact many areas critical to DOE's needs, such as the development of specific biosensors (with nearly single molecule detection limit), and of ultra-dense, ultra-fast molecular-sized electronic components, with very small power requirements and persistent, reprogrammable memories.
For more information, please contact:
William Shelton
|
