'We Did Not Limit the Time for Questions'

© HSE University

The International Laboratory for Supercomputer Atomistic Modelling and Multi-Scale Analysis at HSE University held a major conference on molecular dynamics. Participants had the opportunity to attend all the presentations, while speakers were given as much time as they needed to answer questions. The HSE News Service interviewed Grigory Smirnov, Head of the Laboratory, and Genri Norman, Chief Research Fellow, about the conference preparations and the discussions it generated.

Discussions following the conference presentations, as well as informal conversations on the sidelines, highlighted several promising areas for research, including force-field modelling and predicting the development and interactions of living cells, which could significantly streamline experimental work.

— What were the main goals of the conference this year?

Genri Norman

© HSE University

Genri Norman: The late Richard Feynman, an outstanding American physicist who is considered one of the leading physicists of the twentieth century, would have appreciated this conference. At the California Institute of Technology, Feynman taught a famous introductory physics course that was later published as a three-volume book. The first volume covered mechanics, heat, and molecular physics. In the book, Feynman famously wrote that if, in some cataclysm, all scientific knowledge were to be destroyed and only one sentence containing the most important information could be passed on to the next generation, it would be the atomic hypothesis. This single statement conveys an enormous amount of information about the world: all matter is made up of atoms and molecules that attract one another at a distance and repel when they come close. Although the idea of atoms as indivisible particles dates back to the ancient Greek philosopher Democritus, by Feynman’s time atomic theory had already been transformed by the advances of quantum mechanics and experimental physics.

What has happened over the past 60 years? First, computing technology has advanced rapidly. Sixty years ago, it was still in its infancy. Quantum methods for calculating the interactions between atoms at short distances have also made significant progress, with many of the key advances in calculating short-range repulsive forces coming over the past 25–30 years.

Over the past 30 years, advances in computing and quantum mechanics have enabled researchers to calculate the forces of attraction and repulsion between atoms and to test Feynman’s hypothesis. We can now model how atoms interact and attract one another and use computers to calculate the interactions of large numbers of particles.

When scientists began working on this at the turn of the 1960s and 1970s, we could calculate the interactions of only a few hundred particles. Today, individual projects can involve billions or even trillions of particles. This is the essence of molecular dynamics, which was the main focus of the conference.

— Why did you decide to hold the conference in Pushkin, St Petersburg?

Genri Norman: We initially developed our ideas at a series of symposia on the fundamental principles of atomic modelling, which were held in Abkhazia from 2003 to 2022. Last year, at the initiative of our colleagues, the event was revived in Novosibirsk under a new name and with a broader range of participants. This year, we held the second conference in Pushkin, the famous site of Tsarskoye Selo.

Grigory Smirnov

Grigory Smirnov: The conference was held at HSE University’s Kochubey Centre, which has been operated by the university since 2012. The centre’s classrooms and hotel facilities, all located in the same building, made it possible to accommodate a large number of participants and created favourable conditions for productive scientific discussions.

— How was this year's conference different from previous ones?

Genri Norman: At most conferences, including international ones, there are strict time limits: 20 minutes for a presentation and three minutes for questions. In contrast, we did not limit the time for questions as long as the session generally stayed on schedule. With around 60 presentations in total, we allowed speakers some flexibility, giving more room for interesting discussions. Having plenty of questions and answers during a presentation is a good thing. Participants could then continue discussing topics of particular interest in greater detail during informal conversations on the sidelines, which everyone appreciated.

Grigory Smirnov: We did not have parallel sessions, so participants could attend all the presentations without having to rush from one room to another. This format helps broaden early-career researchers’ horizons and encourages exchange of methods and ideas across different fields.

— What were the main topics discussed at the Molecular Dynamics Conference?

Grigory Smirnov: As the name suggests, the conference is united by a research method rather than a particular subject area. As a result, the range of topics discussed was quite broad. Among them, I would highlight soft-matter physics, which focuses on liquids, polymers, proteins, and other biomolecules; metal physics; and materials science, which includes research on solids and glasses. I would also mention materials under extreme conditions. This is a particularly important area of research, as experimental verification can be difficult, while computational modelling can help expand our knowledge of matter under such conditions.

Genri Norman: One specific example is the Earth’s core, which is subject to extremely high pressures and temperatures. Its properties vary depending on the model used. What is clear is that the solid inner core is surrounded by a liquid outer core. In fact, we live on a planet whose interior remains poorly understood.

Grigory Smirnov: In addition to extreme conditions, one should also mention more everyday problems, such as various biological applications. This is a rapidly developing area, driven in part by advances in computing power. Today, individual properties of biomolecules can be calculated even on a personal computer, and many pharmaceutical companies use such calculations as one of the tools in drug development.

Genri Norman: This is a key direction for the future of molecular dynamics. We are made up of molecules too, so the question is what distinguishes a human being from a stone or, for example, from the keyboard we are using to record this conversation. Physicists have long debated whether there is anything fundamentally different between humans and inanimate objects. I am convinced that there is. Exactly what makes us different remains a subject of research, and so far, there is no clear evidence in the scientific literature beyond the fact that we are made up of moving atoms and molecules. However, according to the laws of dialectics, as the number of interacting atoms and molecules increases and their organisation becomes more complex, new qualities emerge.

This is a topic we hope to explore at our future conferences.

Grigory Smirnov: I would also highlight new methods for data processing, modelling, and analysing various processes. Many presentations focused on these topics as well. Computing power is important, but so are the methods. Machine learning and neural networks are now used both to analyse data and to calculate the forces mentioned by Feynman. These methods were widely represented at the conference.

— Was it a traditional in-person conference, or was it organised in a hybrid format?

Grigory Smirnov: The conference was held in person and brought together 102 participants, most of them from Russia. The international participants included Wang Kai, a professor at Nanjing University; a PhD student from Guatemala; and an employee of HSE University who also works at a Chinese university.

Genri Norman: I would add that Nanjing University is one of China’s top 10 universities. We have established a collaboration with the university, and I hope they will send more participants to the next conference.  

— What trends in the development of molecular dynamics are most noticeable today?

Grigory Smirnov: Molecular dynamics turns 70 this year. In August 1956, scientists from the Livermore Laboratory in the United States presented a paper titled ‘Molecular Dynamics by Electronic Computers’ at a conference in Brussels. Over the decades, computing power has increased dramatically, enabling rapid advances in molecular dynamics methods. We can expect the systems we are able to model to become increasingly complex. If initially we focused on the simplest atomic systems, using model substances to make predictions and interpret them in relation to real-world systems, we can now model increasingly complex systems. These may consist of multiple interacting subsystems, including proteins, parts of living cells, and other complex organic structures. In the case of metals, we can model increasingly large structural units. Finally, greater accuracy has made it possible to develop increasingly sophisticated force fields to describe interactions between particles. As a result, research is moving towards modelling more complex phenomena, making it possible to accurately predict real-world effects.

Molecular dynamics is being used to make quantitative predictions across a range of fields, particularly biology. Researchers are actively applying these methods to develop next-generation drugs and drug-delivery systems, and also to understand how drugs act at the atomic and molecular levels and interact with proteins. This could make molecular dynamics one of the key tools in applied research in the coming decades.

Participants of the Molecular Dynamics Conference

© HSE University

— In which areas are the results of molecular dynamics research most widely applied?

Grigory Smirnov: Molecular dynamics has applications in biology and medicine, particularly in the development of new drugs, as well as in materials science, where an understanding of processes at the atomic level helps researchers build multiscale models and develop new alloys and materials for the space industry and everyday use. Molecular dynamics can also be used to study matter under extreme conditions, such as those found in the Earth’s core, investigate the behaviour of astrophysical objects, and, ultimately, deepen our understanding of the Universe.

— What challenges are researchers in this field facing today?

Grigory Smirnov: The main challenges are related to the increasing complexity of the systems under study and the growing volume of data that needs to be processed. It is no longer enough simply to formulate a problem and run a calculation. Researchers must process the resulting data, extract useful information, interpret it, and use it to develop new theories. At the same time, excessive reliance on computational methods may create certain risks, as researchers may become less critical and start trusting AI-generated results that resemble scientific findings, only to discover upon closer examination that they are unfounded.

The key issues are reproducibility and reliability of results. The molecular dynamics method is based on assumptions related to the choice of force fields. An inappropriate choice can lead to incorrect results, and such errors may not be immediately apparent.

A related challenge is the rapid development of AI and machine learning. On one hand, they can make research more efficient, since molecular dynamics is closely linked to advances in supercomputers and high-performance computing. The more powerful the available computers, the more accurately we can perform simulations and the wider the range of substances and systems we can explore and model.

This development will continue, but the capabilities of modern computers are approaching their limit. Moore’s Law, which predicted transistor counts doubling roughly every 1.5–2 years, no longer holds in the same way. As we have approached the limit in transistor density, future gains are likely to depend increasingly on software optimisation and potentially computers based on fundamentally different principles. There is currently active debate about how quantum computers could affect calculations in our field and how powerful they can ultimately become, but their prospects remain uncertain.

— Is it possible to design a computer specifically for molecular dynamics calculations?

Grigory Smirnov: The development of specialised computers and dedicated computing platforms for particular tasks such as cryptocurrency and data mining is already an established field. There have also been some efforts to develop computers specifically for molecular dynamics. This is a complex and costly approach, but it can produce a computing platform optimised for a particular type of calculation, expanding the possibilities for studying specific systems.

It is difficult to say how in demand or promising this approach will be in the long term, but it is certainly an area that continues to develop.

— Which areas of research, in your opinion, will be the most promising over the next 5–10 years?

Grigory Smirnov: We are increasingly using machine-learning methods in molecular dynamics. Two areas stand out in particular: using machine learning to analyse and structure data and facilitate information retrieval, and developing new interatomic potentials and force fields. For example, when developing a new system, we need to understand how its atoms interact. Machine learning can help us describe these force fields more accurately, making it easier to simulate the system while maintaining a high level of accuracy. The development of universal force fields raises the possibility that, within the next 5–10 years, we may be able to describe the entire Mendeleev periodic table using a single force field—something that would have seemed like science fiction just a few decades ago.

Another promising area is the modelling of living systems, including interactions between proteins and other cellular components. This could allow researchers to simulate processes within a cell or, for example, a virus. Some work in this area is already under way, although largely as a demonstration of what is possible, since such systems are extremely large from the perspective of molecular dynamics. Nevertheless, modelling increasingly complex biological systems looks particularly promising over the next decade.