APS March Meeting 2010
Volume 55, Number 2
Monday–Friday, March 15–19, 2010;
Portland, Oregon
Session P28: Focus Session: New Frontiers in Electronic Structure Theory III
8:00 AM–10:48 AM,
Wednesday, March 17, 2010
Room: C124
Sponsoring
Unit:
DCP
Chair: Martin Head-Gordon, University of California, Berkeley
Abstract ID: BAPS.2010.MAR.P28.1
Abstract: P28.00001 : Advances in Orbital-Free Density Functional Theory: Physics and Algorithms*
8:00 AM–8:36 AM
Preview Abstract
Abstract
Author:
Emily Carter
(Princeton University)
Orbital-free density functional theory (OFDFT) is a first
principles quantum
mechanics method that can scale linearly with system size by solving
directly for the electron density instead of introducing an
auxiliary set of
one-electron orbitals as is done in conventional Kohn-Sham (KS) DFT.
Orbitals must be kept orthonormal; imposing this constraint
involves a
cubically scaling step. KSDFT can be made to scale linearly beyond a
crossover point within a localized orbital framework and hence
molecules and
insulators can be made to scale linearly within KSDFT. Metals
generally do
not exhibit linear scaling within KSDFT due to their inherently
delocalized
electronic structure. OFDFT offers an alternative scheme by
introducing a
kinetic energy density functional (KEDF) and local electron-ion
pseudopotentials (LPSs). Thus the accuracy of OFDFT depends on the
representation of these two terms. We now have a routine tool for
constructing accurate LPSs by inverting the KS equations for bulk
crystals.
These BLPSs are validated against accurate nonlocal PSs within
KSDFT. A
decade ago, we reported a nonlocal KEDF that accurately captures
the physics
of nearly-free-electron-like metals. Here we report a new
nonlocal KEDF that
explicitly contains the physics required to describe semiconductors.
Extensive tests on silicon and a variety of compound
semiconductors reveal
this new KEDF to be accurate for many properties, thus opening up
the door
to accurate OFDFT calculations on semiconductors. We also report the
systematic elimination of bottlenecks within our OFDFT code that
render the
entire algorithm linear scaling for all system sizes (no
crossover point).
With parallelization then introduced via domain decomposition,
quantum
mechanical simulations of metal samples containing up to 1
million atoms
have been demonstrated using a modest number of processors. With
this new
capability we are studying mesoscale features that control
mechanical
properties of Al and Mg alloys.
*National Science Foundation
To cite this abstract, use the following reference: http://meetings.aps.org/link/BAPS.2010.MAR.P28.1