SciDAC Institutes SciDAC Partnerships Basic Energy Sciences Biological and Environmental Research Fusion Energy Sciences High Energy Physics Nuclear Physics Participating Orgs SciDAC Overview (2000-2011) Logos DOE Office of Advanced Scientific Computing Research (ASCR)

New Coupled-Cluster Methods for Molecular Potential Energy Surfaces

PI:

Piotr Piecuch, Department of Chemistry, Michigan State University

Affiliated Researchers (SciDAC):

• Professor Mark S. Gordon, Iowa State University and Ames Laboratory
• Dr. Michael W. Schmidt, Associate Scientist, Ames Laboratory

Interactions with Other SciDAC Participants:

• Dr. Robert J. Harrison, Oak Ridge National Laboratory
• Dr. David J. Dean, Oak Ridge National Laboratory

Other Collaborators:

• Professor Stanislaw A. Kucharski, University of Silesia, Poland
• Dr. Monika Musial, University of Silesia, Poland

Summary

This research focuses on the development of new generations of electronic structure methods and efficient general-purpose computer codes that can provide a highly accurate description of chemical bond breaking, reactive pathways, and electronic excitations in molecules. The new coupled-cluster methods developed in this program will enable precise modeling of elementary chemical reactions that occur in combustion, catalysis, photochemistry, and photobiology using the state-of-the-art multi-processor high-performance computer platforms.

Background Information

Molecular electronic structure calculations, followed by dynamical studies, have been recognized as a cornerstone for the successful modeling of chemical reactions occurring in combustion, catalysis, and photochemistry. New ab initio approaches, based on the first principles of quantum mechanics, which can accurately describe the complicated motions of electrons in molecules (the electron correlation effects), are critical for these developments. Of particular significance are the approaches that can reliably predict entire potential energy surfaces of ground and excited states of molecular systems, since potential energy surfaces determine how atoms and chemical bonds rearrange during chemical reactions. To meet the challenge created by new generations of high-performance computers and to advance ab initio theory to a new level of accuracy and applicability, so that reliable calculations for all kinds of molecular systems and electronic states become routine, the efficiency of the existing computer codes has to be increased and new generations of electronic structure methods and algorithms must be developed.

Principle Goals

This research focuses on the "holy grail" of the ab initio electronic structure theory, which is the development of "black-box" and yet very accurate coupled-cluster (CC) methods and computer codes that can provide an excellent description of bond breaking, reaction intermediates, molecular potential energy surfaces, and excited electronic states. Our new CC approaches, termed the renormalized coupled-cluster methods, the active-space coupled-cluster methods, and the method of moments of coupled-cluster equations (MMCC) are capable of describing entire molecular potential energy surfaces and excited states at the fraction of the effort associated with the more traditional multi-reference configuration interaction (MRCI) calculations. Thus, our new methods can be applied to much larger systems and basis sets, and new chemical problems that could not be handled by the existing CC and MRCI approaches. We can already apply our new methods to systems consisting of 20-30 light atoms, clusters of ~10 transition metal atoms, and hundreds of basis functions. Our plans to develop parallel CC codes will enable calculations for even more complex systems. All of our CC computer codes are fully vectorized and can access large (many GBs) memories available on the state-of-the-art computer platforms. Our CC codes are shared with the entire community by incorporating them in GAMESS, which is a highly scalable electronic structure package distributed at no cost by Professor Mark S. Gordon at Iowa State University and Ames Laboratory. Currently, GAMESS has nearly 9,000 registered users. Thus, by incorporating our CC codes in GAMESS, we provide many scientists with powerful new research tools that have not been available to them before. development and demonstration of CSP mechanism reduction components.

The Most Important Accomplishments (September 1, 2001-January 31, 2003):

1. The incorporation of the standard closed-shell CC approaches (LCCD, CCD, CCSD, CCSD[T], and CCSD(T)), and new and highly promising renormalized CCSD[T] and CCSD(T) methods for single bond breaking in the GAMESS package. The initial implementation, released in June 2002, has been followed by an improved version (developed in January 2003), which provides a better memory/disk management, enabling routine calculations for ~ 100 correlated electrons (hundreds of electrons total) and hundreds of basis functions.
2. The MMCC and renormalized CC methods have been extended to excited electronic states (December 2002). These new "black-box"methods provide ~ 0.1 eV accuracies for a wide range of excited states and larger (10-20 atoms) molecular systems.
3. The renormalized CC and MMCC methods have been extended to multiple bond breaking by developing the quadratic MMCC, CI-corrected MMCC, and extended CC methods (June 2002-January 2003).
4. Strong evidence has been provided that the exact many-electron wave functions can be represented by cluster expansions employing two-body operators. This result may have a considerable impact on all calculations for many-body systems.

Specific Plans for 2003-04:

1. GAMESS will be substantially enhanced by incorporating in it the equation-of-motion CC (EOMCC), MMCC, and renormalized CC methods for excited electronic states.
2. The efficient codes for the new MMCC methods for multiple bond breaking will be developed.
3. Open-shell extensions of the CC/EOMCC and new MMCC, renormalized CC, and active-space CC methods will be developed.
4. Initial efforts will be made to parallelize the CC codes already in GAMESS (in collaboration with Professor Mark S. Gordon and co-workers).

Impact of the SciDAC Team Effort:

The SciDAC team approach has greatly enhanced the way the PI’s group develops new computer codes. SciDAC has allowed us to collaborate with Professor Mark S. Gordon and Dr. Michael W. Schmidt and to incorporate our CC codes in GAMESS. Our codes have benefited from the software tools and excellent scalability and memory management provided by GAMESS. Our interactions with other SciDAC participants have been fruitful, too. Dr. Robert J. Harrison (ORNL) helped us with parallel eigensolvers. We also collaborate with Dr. David J. Dean (ORNL) on applying our CC codes to nuclear structure. None of this would ever happen without SciDAC.

Computational Needs:

In order to apply our codes to larger molecular problems relevant to combustion, photochemistry, and catalysis, we would need 200,000 node hours/year on the NERSC systems.

back to project page