A Catalyst for Discoveries

Michele Parrinello, the father of many methods that have paved the way to understanding some of matter’s deepest secrets, has just turned 80. A few days after his birthday, he was at ICTP—of whose community he is a cherished member—to attend the conference “Frontiers in Atomistic Simulations: From Physics to Chemistry and Biology,” surrounded by a large community of researchers, many of whom have been his collaborators.

In this interview, he talks about his life’s work, which has influenced and inspired many scientists around the world.

The Parrinello-Rahman method, the Car-Parrinello method, metadynamics, the Behler-Parrinello neural networks - these are just a few of your many contributions to our understanding of physics and chemistry, and you are still working on several topics. What drives you?

I have always been motivated by the will to solve real problems in physics and chemistry and I am very passionate about what I do. The methods are just a tool to get there. I like solving problems and working on these simulations is something that continues to fascinate me. I also love working with young people. Many among them have a special type of intelligence and it is nice to be with them. Without that, do you think I would still be working at 80?

You started working on simulations at a time when theoretical physics was considered the best option for a good student like you. How did you end up in this field?

It all happened by chance. I started with relativistic field theory for my final project at the University of Bologna. The professor I worked with was a great lecturer, but he was not around very much, so I had to work on the project on my own. It was hard, but it also taught me that I could make it through such difficult situations. After that, I met Mario Tosi, who was doing some interesting work in Messina, where I am from, and I got a chance to continue my research there. My girlfriend at the time was also living in Messina, so my motivations to move there were not only scientific. Everything else came from there. I can say that my start in research was not great, but finally things ended up well for me.

Which one of your many contributions are you most attached to?

I would say the Parrinello-Rahman method, because it was the first one, and also because it was done in collaboration with Aneesur Rahman, who was like a father to me during the time that I spent in Chicago. It is he who taught me everything in molecular dynamics. He helped both me and my family at a time when I was not well. Because of all these factors, that contribution is the one that I feel most attached to. This said, Roberto Car is like a brother to me and I cannot forget the excitement we shared when we were working here together.

You and Roberto Car developed the famous Car-Parrinello method in Trieste, in the hallway of the ICTP Leonardo building. What do you remember of those days?

We were very enthusiastic and worked days and nights. Because the computers were less busy at night, we used to come back to work in the evening, after everyone else had left. In order to stay awake, on the way here we used to grab something to eat at the California Inn [a diner close to ICTP that is now closed]. We looked like two nerds surrounded by all this beautiful youth, the ‘mule’ and ‘muli’, as they say here in Trieste, who were having fun while we were completely absorbed by our simulations.

If your life was a chemical process, what would be the catalysts of your discoveries?

My thesis supervisor was a negative catalyst, but I had at least three positive ones. Rahman, whom I have already mentioned, and then also Erio Tosatti. It is he who brought me to Trieste and I owe him a lot. He passed his enthusiasm on to us and he showed us how to work collectively, as a team. He has been a fundamental person in my life. The other one is Roberto Car, of course.

You were a pioneer in the use of Artificial Intelligence. With Behler you understood that machine learning could help your simulations already twenty years ago. How did that come about?

The Car-Parrinello method was very powerful, but also slow. People had tried several ways to accelerate it and avoid calculating the potential at each time step, but always with some drawbacks. With Behler we were not the first to think that using machine learning to represent the potential could be useful, but we were the first ones to couple it with molecular dynamics. That also implied taking care of some technical details to ensure that the potential respects all the right symmetries. The best part was to see that it worked when we applied it to a real system, silicon.

Over the years you have supported ICTP’s mission in many ways, particularly by serving on its Scientific Council for many years. What would you say to scientists in developing countries who want to do simulations but do not have access to the computational power needed to carry them out?

Those material conditions are of course important, but wit goes a long way. If one is ingenious and has the will to work on something, they will find ways to do good science.

How did you like the conference on “Frontiers in Atomistic Simulations” so far?

It has been great. Of course, I am biased, because most of the people here are somehow my children. But setting aside the more emotional aspects, the conference has been a display of incredible contributions that, five or ten years ago, would not have been possible. It has beautifully showcased all the scientific progress that has been done in this field.

What’s a development that you think will come soon and that you look forward to?

As someone said, I can predict many things but not the future, so I do not know. It is hard to tell because there are many factors that come into play. For example, machine learning did not exist before and who would have predicted all the possibilities that it has opened up? Of course, if quantum computing becomes a reality, it will change many things.

But I hope that also the methods that we have developed so far will encourage more people to carry out their research and solve more exciting problems.

Michele Parrinello talked about some of his most recent contributions and his fascinating life in science in his colloquium on “The Weird and Wonderful World of Catalysis and How I Got There”:

Watch Michele Parrinello in conversation with Atish Dabholkar, ICTP Director, and Sandro Scandolo, ICTP Head of Research, as part of the Dirac Conversation video series:

Photos of Parrinello’s colloquium can be found here.