- American Physical Society Sites
- Meetings & Events
- Policy & Advocacy
- Careers In Physics
- About APS
- Become a Member
This is the sixth in a series of articles by James Riordon. The first article appeared in the November 2002 issue.
"It was like a short blanket," says Michele Parrinello as he explains the hodgepodge of theoretical methods available in the early '80s for describing the respective crystalline and liquid phases of silicon. "You know, you pull it over your head and your feet stick out in the cold, and vice versa." At the time, Parrinello was a professor at the International School for Advanced Studies (SISSA) in Trieste, with an interest in the phase transition of melting silicon. "It was a popular subject at the time," continues Parrinello, "Silicon is a semiconductor. It is well known in the crystalline phase. And when you melt it, which happens at about 1700k, it goes into a metallic state. People were trying very hard to devise potentials that would be suitable for the solid state phase and for the liquid state phase."
Parrinello and SISSA colleague Roberto Car decided that computer technology had advanced sufficiently that they might be able to devise a number of tricks to efficiently model the phase transition via ab-initio molecular dynamics. Density functional theory allowed them to replace the many-body electronic structure problem with an effective single particle problem. To put it somewhat simplistically, once they calculated a three-dimensional electron density, they could determine how the molecules would respond to the distribution. The new molecular arrangement would lead to a new electron density, and so on. In hindsight, the algorithm seems fairly straightforward, but as Car points out, "It was at that time considered simply not possible to use that kind of theory to base a real molecular dynamics simulation."
The unified scheme combining molecular dynamics and density functional theory, commonly called the Car-Parrinello algorithm, ultimately turned out to provide plenty of blanket to keep the researchers' toes and ears simultaneously warm, so to speak. "We had this idea that it could be useful in a variety of contexts," says Parrinello, "but what is amazing is that same code?without changing anything but the atomic number?can model ionic systems, semiconductors, metals, hydrogen bonds, water, hydrogen fluoride, protein. You see what I mean? It is the same code that can do this enormous range of applications."
"I think what was very good," says Parrinello, "was Roberto had the background in electronic structure theory, and my background was more on the side of molecular dynamics and statistical mechanics, and we could talk well to each other. Once we decided to start with exploring the possibility of calculating things abinitio it all went very fast. I mean, the code and the first results came in a few months."
The Car-Parrinello algorithm is now a standard tool in condensed matter physics, but it has perhaps been even more useful in other disciplines?a fact that is reflected in the many awards that the two have garnered. Although Parrinello and Car shared the 1995 APS Rahman Prize for Computational Physics, says Parrinello, "I personally receive more recognition from the chemists than from the physicists. I don't have training as a chemist, but most of my papers tend to go into chemical journals."
Both researchers have published dozens of papers extending their algorithm to new regimes, including the biological sciences. Parrinello is particularly interested in studying the nuances of water. "Water is interesting for two reasons; first it's a very subtle system so it's a challenge for the theory, and the other is that water is a solvent for many chemical reactions, for electrochemistry. And water is of course necessary for life, so we are trying to build toward biophysics and biochemistry. So that's why we are focused on water, because it is such an important substance for life," says Parrinello.
In recent years, Car has broadened his research to include such things as modeling electronic current in nanostructures, in addition to his continuing work on extending the Car-Parrinello algorithm. But regardless of the particular problem he is addressing, Car feels that he is still benefiting from the collaboration with Parrinello. "The way we approached the problem when we came out with the Car-Parrinello approach has given a permanent inspiration to all my subsequent activity," says Car. It has also opened his eyes to problems that he might not otherwise have recognized. "When you make some important progress in one field, this always make you think more deeply about fundamental problems," Car explains, "and you inevitably discover that in spite of your progress you did not solve all of them."
Parrinello has experienced similar benefits from the highly-cited work. "I've not been afraid of new things and new areas and new problems. That's probably my greatest virtue. To move from one subject to the other to face various new challenges. It's important because often people don't move because of fear. I'm always growing, as it were. I didn't start like that, but growing and having success with this paper and a couple of other things gave me courage to face the world."
Parrinello is now Director of the Swiss Center for Scientific Computing and Professor of Computational Science at the Swiss Federal Institute of Technology in Zurich. Car is a professor in the Chemistry department and at the Princeton Materials Institute is affiliated with the Physics department and the program in applied math. The two researchers have not worked together since their brief collaboration on the Car-Parrinello algorithm, but they continue to share awards for their method and maintain a strong personal bond. "We are brotherly friends on very good terms," says Parrinello, "but the dynamics of life have taken us to different parts of the world and different interests." And neither they nor the countless scientists who apply the algorithm must suffer with the short blanket that once hampered studies of the molecular dynamics of condensed matter phase transitions.
©1995 - 2022, AMERICAN PHYSICAL SOCIETY
APS encourages the redistribution of the materials included in this newspaper provided that attribution to the source is noted and the materials are not truncated or changed.