PNNL-13501 - Pacific Northwest National Laboratory

Study Control Number: PN98013/1259
Combustion Simulation and Modeling
Jeffrey A. Nichols, Robert J. Harrison
High-performance calculations require efficient computer algorithms to achieve efficient modeling of the
thermodynamics and kinetics of gasoline and diesel fuel combustion. Combustion simulation and modeling have
important practical applications for predicting reaction rates and products.
Project Description
The purpose of this project is to develop highperformance
computational chemistry software for
massively parallel processor systems in the area of
coupled cluster (CCSD(T)) methods for treating the
electron correlation problem. Coupled cluster methods
are needed to achieve the accuracies required to replace
experimental measurements for many areas of chemistry,
but especially for the combustion simulation and
modeling community. Unfortunately, the coupled cluster
method scales as N 7 , where N is the number of atoms in
the system. This greatly limits the range of applicability
of this method to combustion and other problems. The
project focuses on developing methodological and
computational approaches to bring about a dramatic
reduction in the computational cost of coupled cluster
calculations. This entails algorithm redesign with the use
of new methods in order to achieve the reduced scaling
and increased massively parallel processor efficiencies
needed to accurately model the thermodynamics and
kinetics of large molecular systems such as those found in
gasoline (C8) and diesel fuel (C16) combustion.
This work is accomplished within the framework of
NWChem (Kendall et al. 2000), a computational
chemistry suite that is part of the Molecular Sciences
Computing Facility in the Environmental Molecular
Sciences Laboratory. In addition, POLYRATE (a
chemical dynamics package developed at the University
of Minnesota, in collaboration with PNNL staff, and used
to predict rate constants) (Steckler et al. 1995; Corchado
et al. 1998) will be integrated with NWChem in order to
provide improved means for the calculation of chemical
reaction rate constants.
Results and Accomplishments
The integration of POLYRATE with NWChem to
perform direct dynamics has been completed and will be
widely distributed in NWChem version 4.0, which is due
for release in fall 2000. It is already available for
Laboratory internal use and is being used by the
atmospheric oxidation study.
Alternative representations of the Møller-Plesset
perturbation theory and coupled cluster wave functions
have been explored with some limited success. Our goals
were to develop a framework to improve the scaling of
high-accuracy calculations in large atomic orbital basis
sets by using a low-rank representation of both the wave
function (coupled-cluster amplitudes) and integrals.
Our investigation into the Laplace-factorized algorithm
for the 4 th -order linear triples contribution is continuing in
collaboration with Professor J.H. van Lenthe, from U.
Utrecht, who visited this summer.
Most of the effort this fiscal year has been devoted to the
new production code for CCSD(T). Several prototypes
were finished earlier this year, with the objectives of
validating the equations and exploring various methods.
Significant highlights of the new CCSD(T) code include
the following:
• Hartree-Focks RHF/ROHF and UHF reference
functions.
• Full use of Abelian point-group symmetry.
• Design goals set by considering calculations on
molecules using cc-pVnZ (n=2-6) basis sets (also
with augmentation) that could be executed in one
week on a 1 TFLOP/s computer with 1 gigabyte of
Computational Science and Engineering 109