Interview by Carey Sargent, EPFL, NCCR MARVEL, November 2022
Have you always been interested in science?
It’s an interesting question, and the answer is yes. Of course, a lot of the credit for that goes to my math and physics teachers in middle and high school. But I made my choice about a week before I had to enrol at university. The alternative would have been history. This was much more in the landscape of my family—they are all from the humanities. I remember having a wonderful chat with my dad, who was a historian by training, and he told me “look, I think you can read a professional history book even if you do not have a degree, but physics? It’s not quite the same, so go for it.” And that’s how I decided. For me, as a teenager or a young person, I never had a notion that this would be weird. But maybe when I arrived at university and saw how few women were enrolled in physics, that struck me, and I thought maybe I had made an unusual choice. And of course, today, in my professional life, that is one of the themes.
Why are you a scientist?
It’s a combination of two things. Curiosity, of course, but systematized curiosity in some sense. Ignorance is a key driving force in the sense that the older I grow, the more I change my relationship with ignorance. Maybe when you’re a young scientist you come into the game thinking “I want to understand lots of things” and then you realize that, really, the questions sometimes are much more fun than the answers. That’s one aspect.
I also love the fact that, particularly in physics, we never, or very rarely deal with exact theories or exact results and so we live with approximation that you can quantify, or call ignorance if you are in a more poetic mood. The whole game is how much can you learn within the boundaries of uncertainty and approximation.
So, when I joke with my friends, I say physics is not an exact science, it’s a science of controlled approximation. That I love, it is endlessly fascinating: how much we can do when we understand, accept and play the game in this way. Airplanes fly, we can devise new materials, we can understand biology to some extent, accepting that our knowledge is within boundaries. By pushing these boundaries, we make contributions, but there’s always going to be a new boundary.
How did you become interested in modelling and simulation?
I basically fell in love with simulations during my third year of university, when, by chance, I had a gap in my timetable of classes. A friend of mine, who knew that I loved statistical mechanics, suggested that I might be interested in at least auditing a class where there was a fussy, strict, strange character who used computers to solve problems in statistical mechanics and to simulate systems, materials, matter, etc. etc. So I went and this is how I met Giovanni Ciccotti. I not only audited the class, but I took the exam and really liked it and then chose him as my advisor for the master’s thesis project, which was fantastic. I graduated with him with a thesis on Simulating Quantum Dynamical properties of nuclei.
A lot of what is done in MARVEL has to do with treating the electrons at the fully quantum mechanical level, with ab initio molecular dynamics, for example, but, more often than not, we assume that the nuclei are classical, so Newton rather than Schrodinger. One of the things I was passionate about at the time and this remains a theme of my research even today is how to go beyond that. This is a super difficult problem because the costs of simulating the dynamical quantum properties scale exponentially with the number of degrees of freedom. Quantum computers may one day help us to make progress there, but for the time being there is no way that we can get an exact solution for a system of any respectable size.
There is a lot of research—Michele Ceriotti, for example, is someone else within the MARVEL community who deals with these sorts of things and a lot of research goes into finding a reasonable, accurate, controllable, but numerically efficient approximations to incorporate these quantum properties, which are very important. In particular for materials that contain hydrogen, but also lithium, anything light, or in the appropriate temperature or pressure conditions, they can affect the properties of a system enormously. So if you’re interested in understanding the spectrum of a system that contains water, you need that. If you want to understand the diffusion of light of materials, you need that. If you want to understand the energetics or the free energy of formation of certain materials, especially when they include hydrogen, it’s something you need. That’s what I did for my master’s thesis and then I continued doing that during my PhD, when I moved to Boston, where I was at Boston University for 3.5 years.
What’s the aim of your research in general?
What I said about quantum dynamical effects is where it all started. I would say that by training and by passion my focus is mostly on the development of methods and new algorithms to try to solve problems more efficiently or more generally. It’s not really applications, it’s more problem-driven in the abstract sense. In that sense, I would say that I have benefitted from a very broad education and, afterwards, research topics.
Quantum dynamics is certainly one of the core bits, but I would say that in general statistical properties of condensed state systems is my broad area, there’s also a lot of statistical mechanics, which means activated events, systems that have very different time scales involved in the motions of their constituents. This is at the abstract level, then applications come a little bit with that. For example, I have been engaged in research that had to do with materials for solid state hydrogen storage.
What I am going to do in MARVEL in the beginning at least is related to my method development passion…married with the idea of proposing a new practical tool to the community. We have a new way to integrate the Born-Oppenheimer equations of ions and electrons in the ab initio molecular dynamics that we call Mass-Zero Constrained Dynamics, or MaZe for short. The core of my activity with MARVEL will be to code that in a library in collaboration with CSCS and Giovanni Pizzi so then we can efficiently use these codes in the electronic structure and materials community. The reason we’re interested in doing this is because preliminary indications point to the fact that this way could be more efficient and solve some of the conceptual limitations of alternative methods. We have to explore that to see how much this is going to be true.
What motivates you?
In addition to what we’ve discussed, there is another aspect of my current job, which combines active research with my duties as Deputy Director of Centre Européen de Calcul Atomique et Moléculaire (CECAM). CECAM is the oldest organization for the promotion and development of simulation and modelling. It was created in 1969 in Paris and the headquarters have been at EPFL since 2008. CECAM is a broad, international European-level institution funded by 25 member organizations in Europe and we have one Chinese membership. There are 17 CECAM nodes across Europe where we have activities like workshops and conferences, but also research projects, training, networking, etc. etc.
The reason that I mention this is because we have all benefitted from institutions and infrastructures and people as we were growing up as scientists, I feel that it is important to give something back, to be of service to the community in some way. Also, it’s interesting because it opens challenges that are more related to scientific strategies, to creating infrastructures for the community, and I think that what we’re going to do in MARVEL with the simulation as a service model goes very much in that direction. It’s also challenging in unusual ways because, as with everything that puts together more than four human beings, and with this being a European institution…sometimes it’s like the joke, “An Italian, a Frenchman and an Englishman go into a bar…”
It’s also nice because it enables to me to somehow push the boundaries of what I try to do. So, for example, we have this collaboration with MARVEL on outreach, which is going to be even more important in phase 3. CECAM and MARVEL co-organize a joint lecture series, joint online events, and we also have something that I am particularly fond of, the Mary Ann Mansigh Conversation series. These really enable us to have conversations with people who have played a role in our community in the very broad sense. Early programmers, people who developed early machines, people who work in industry…but they’re not technical seminars on science. It’s more of a conversation to broaden the horizon. I think this is super important. Dissemination to the general public is also something that I like and I have been involved in creating two comics, one on simulation and one called Cherchez, les femmes to promote female participation.
Which papers are you most proud of?
Always my last one, of course! Some of the latest things that we are doing in method developments, this MaZe algorithm, is something that I am quite proud of because of what I said before
And then, as a PhD, I was working on new methods to talk about what we call non-adiabatic dynamics for mixed quantum-classical systems. This is when, if you have a molecule, if you leave it to its own devices it will vibrate around its minimum energy configurations, the electrons will mostly be in their minimum energy state. Then if you shine a laser on this molecule, you can excite the electrons and, by following the way in which the electrons reorganize themselves after this excitation, you can make chemistry, or you can make physics because you can break bonds, form bonds, this sort of stuff.
There were a couple of papers, one was a PNAS (Proceedings of the National Academy of Sciences) paper in the early 2000s in which, with David Coker, we proposed a new method which was then picked up and used by other people. This was called the LAND-map method and I know that a lot of the work that David has done in the following years in his group is based on those themes. Of course, these are technical, it’s not like Car-Parrinello molecular dynamics, which is a much broader tool, but I think that in my own little way it was a significant contribution.