Supercomputing Institute Fall 2000 Bulletin Vol. 17 No. 1

A workshop to honor the originators of the Car-Parrinello Method and to review major accomplishments and cutting-edge developments within the general area of molecular dynamical simulations was held March 18-19, 2000, at the University of Minnesota Supercomputing Institute for Digital Simulation and Advanced Computation.

The Car-Parrinello method, introduced in 1985 by Roberto Car (Princeton University) and Michele Parrinello (Max Planck Institute), has dramatically influenced electronic structure calculations for solids, liquids, and molecules, and initiated the field of quantum molecular dynamics among physicists. Molecular dynamics is a simulation method wherein the trajectories of individual atoms or molecules are integrated, usually using Newtonian dynamics. These simulations have provided us with great insights into the dynamical and structural properties of matter, such as the microstructure of the liquid state and the diffusion of impurities in crystals.

A key ingredient in this formulation is the description of interatomic forces. Prior to the invention of the Car-Parrinello algorithm, one typically constructed interatomic potentials either from experimental data or from other indirect theoretical methods. These classical methods often worked quite well for some systems (e.g., inert gases), but were not generally reliable for systems with complex bonds (e.g., covalent or partially covalent bonds where the interatomic potentials are inherently many-body ones).

The Car-Parrinello method deals directly with a quantum description of the electronic structure for the system of interest. In the Car-Parrinello algorithm, the electronic degrees of freedom are simultaneously evolved quantum mechanically along with the nuclear degrees of freedom, which are treated by a classical description. Within the Car-Parrinello algorithm the resulting interatomic interactions are fully quantum mechanical in nature-no ad hoc assumptions need be invoked.

The Car-Parrinello method has led to an explosive growth and interest in accurate modeling and simulating a wide variety of condensed matter systems. Applications of this approach can be performed on liquids (both metals and insulators), clusters, disordered solids, biological systems, defects, and surfaces. It has also led to the development of a number of related algorithms for handling large and complex systems.

Roberto Car and Michele Parrinello have both received a number of awards and recognition for their work; for example, the 1995 American Physical Society Aneesur Rahman Prize for Computational Physics was awarded to Car and Parrinello for the development of their algorithm. Aneesur Rahman, for whom the prize is named, was instrumental in the development of the molecular dynamics method. He performed the first realistic molecular dynamics computations and was a professor of physics at the University of Minnesota in the 1980s.

The Supercomputing Institute and IBM sponsored this meeting and it was endorsed by the American Physical Society. Over 80 distinguished scientists attended the workshop, including 24 from 11 countries outside the United States.

The meeting began with a plenary talk by Roberto Car. He spoke on using simulation methods for optimizing multivariable functions. This is a generic, fundamental problem in many fields. For instance, one might want to minimize a complex energy functional by varying a large number of parameters. These parameters might correspond to both electronic and nuclear degrees of freedom. Car outlined a new and novel approach using what he called quantum annealing.

In the following talks, Marvin Cohen (University of California at Berkeley) spoke about the contributions of pseudopotential theory to ab initio molecular dynamics. In particular, he noted that pseudopotential approach allowed one to focus only on the chemically active electrons and made it possible to compute accurate interatomic forces. This observation established the starting point for the Car-Parrinello algorithm. Giulia Galli (Lawrence Livermore National Laboratory) spoke about recent ab initio simulations of water under high pressure. This work extracted unexpected and complex microstructures present in a relatively simple liquid. Leonard Kleinman (University of Texas at Austin) followed this theme by outlining the properties of liquid metals such as sodium obtained using ab initio molecular dynamics. Michiel Sprik (Cambridge University) used a similar technique to examine reactions in aqueous solutions. The last speaker in the morning session was Paolo Carloni (International School for Advanced Studies, Trieste). He presented an overview on recent applications of the Car-Parrinello algorithm to understand complex biological molecules. A long-range application of this work would involve drug design and gene therapy.

In the afternoon session, Sokrates Pantelides (Vanderbilt University) discussed extensions of the Car-Parrinello method to excited state systems, applications which can be observed in the absorption of light and the propagation of electronic current through micromolecules, among other examples. Jerzy Bernholc (North Carolina State University) followed his talk, discussing simulations of nanotubes and their mechanical properties using molecular dynamics. Nanotubes are microscopic tubes of carbon that have novel electronic and structural properties. Chris Van de Walle (Xerox) gave an overview of defects and impurities in semiconductors such as gallium nitride. These systems are highly relevant for work within the electronics industry. Ursula Röthlisberger (ETH Zentrum, Switzerland) gave examples of Car-Parrinello simulations in the area of biochemistry, for example, the study of enzymatic reactions. Her talk also illustrated how one could combine molecular mechanical descriptions of these systems with quantum forces. J. Woods Halley (University of Minnesota) concluded the session by discussing water interacting with metal surfaces. This is an important system for electrochemical processes.

Michele Parrinello started the second day of the meeting with a plenary talk on spectroscopic properties of disordered and liquid systems using Car-Parrinello molecular dynamics. He addressed fundamental issues involved in predicting the low frequency response functions for amorphous semiconductors and water. The following talks focussed on silica and related materials. Don R. Hamann (Lucent, Bell Labs) outlined a computational method based on adaptive coordinates for examining the electronic and structural properties of pure and doped silica. Peter Blöchl (IBM Zürich Research Laboratory, Switzerland) spoke on zeolites. Zeolites are complex silicates used for catalytic processing as seen in the petroleum industry. He illustrated the Car-Parrinello method by considering methanol and water in zeolitic systems. Alfredo Pasquarello (Institut Romand de Recherche Numerique en Physique des Materiaux, Switzerland) talked about the silicon-silica interface, which plays a crucial role in all electronic devices using silicon wafers.

The morning session ended with talks on materials of interest in earth science and geology, as well as the behavior of matter under extreme conditions. Renata Wentzcovitch (University of Minnesota) outlined how ab initio molecular dynamics can be used to predict the elasticity of materials as a function of pressure and temperature. Her work has direct relevance for describing the earth's interior and constructing seismological models. (See article on p. 6.) Sandro Scandolo (International School for Advanced Studies also considered simulations to describe the behavior of matter under pressure and temperature; however, his work focused on non-terrestrial systems such as the planet Neptune.

The afternoon session was comprised of four talks. Richard Martin (University of Illinois at Champaign-Urbana) presented the first, examining methods for improving the efficiency of the Car-Parrinello method to large systems. Most quantum calculations scale in time as some power of the number of atoms considered. For example, if there are N atoms in the system, the computational load might typically scale as N3 for a quantum simulation, whereas a classical simulation would scale linearly with N. Martin's work focused on developing algorithms for quantum systems that scale as classical systems. David Vanderbilt (Rutgers University) presented some simulations for surface dynamics. In particular, he illustrated how Al atoms move on the surface of the metal. Matthias Scheffler (Max Planck Institute) also examined surface diffusion for semiconductors such as GaAs and related growth methods for these systems. Wanda Andreoni (IBM Zürich Research Laboratory, Switzerland) presented the last talk of the meeting. She used Car-Parrinello simulations to examine the optical and electronic properties of complex inorganic and organic salts on metal surfaces such as gold.

	URL: http://
This page last modified on
Website related questions or problems should be directed to webmaster@msi.umn.edu The Supercomputing Institute does not collect personal information on visitors to our website. For the University of Minnesota policy, see www.privacy.umn.edu.