An explorer of chemical space.
Text: Christoph Dieffenbacher
Scientists are exploring the almost endless world of possible molecules and compounds in their search for useful substances. Anatole von Lilienfeld, Professor of Physical Chemistry, is developing new and faster methods in this quest based on quantum mechanics and machine learning.
The old facades of Klingelbergstrasse can be deceiving – this is actually the scene of cutting-edge scientific research. In the almost century-old building of the “Physico-Chemical Institute,” you won’t see laboratories full of flasks, Bunsen burners or measuring instruments. Here, the work is performed almost entirely on computers – except these computers are many times more powerful than their everyday cousins. Researchers use the formidable machines to develop and test new methods for seeking out hitherto unknown substances.
Chalkboards full of formulas
Anatole von Lilienfeld, Professor of Physical Chemistry, welcomes us into his office after he has finished speaking to two of his doctoral students. With a warm face and soft features, he speaks in a deep voice and is wearing jeans with a plain shirt. In addition to a large computer monitor, there is a comfortable sofa, as well as armchairs and laboratory chairs. On the floor, a cozy rug. The shelves contain just a few books along with a couple of photos of his children. Next to them, an object made of sculpted glass – an award he received in the USA.
Perhaps the most striking feature are the chalkboards mounted on two of the walls. Teeming with numbers, symbols and formulas scribbled in chalk, these boards are used to discuss theories and plow through statistics. The professor glances over at them from time to time. “Our aim is to gain a better understanding of chemical space,” he says. “By that, we mean the huge virtual expanse within which all possible compounds exist.” Only a tiny part of this is known to scientists.
To illustrate the point, von Lilienfeld makes a comparison with human DNA: Although it contains sequences from just four bases, these are sufficient to account for the full breadth of biological diversity. By analogy, the hundred or so chemical elements can theoretically be combined in space – in every possible geometric variation – to produce “an incredibly large number” of potential compounds. This has been estimated at 1060 – more than the number of atoms in the entire solar system – and that is just for organic and medium-sized substances.
What things are made of
Raised in Germany as the son of a physician and a theologian, von Lilienfeld says that, even as a child, he was fascinated by what things are made of. Not satisfied with the usual answer, “atoms”, he wanted to know more. Later, he found himself drawn to chemistry because of its widespread applications. Von Lilienfeld is convinced that “even today, we are a long way from realizing its full potential.”
He became a convert to physical chemistry at ETH Zurich, where the supervisor of his dissertation was engaged in a theoretical and experimental study of the quantum mechanical behavior of molecules. After completing his doctoral thesis at ETH Lausanne, von Lilienfeld spent eight years in the USA, where he conducted research into chemical space at UCLA (California), New York University and the National Laboratories in Albuquerque (New Mexico) and Argonne (Illinois).
Today, he lives with his family in the heart of central Basel and finds the small-scale, urban environment very agreeable: “In the USA, I spent hours and hours in the car every week. Here, I can simply walk to work.” He also has a high regard for the working conditions in Switzerland, which he believes are superior to those in other countries. Although he works exclusively on basic research, he is also in regular contact with industry.
Exponential growth
So, what does a scientist get up to in chemical space? Von Lilienfeld and his team investigate the concept using quantum mechanical methods. Their innovative and promising approach is even drawing international acclaim and involves researching aspects such as the properties, distribution and behavior of electrons in molecules. Embedding in chemical space then allows the researchers to more rapidly track down interesting relationships and laws – like the periodic table but with more dimensions.
Of course, the calculations involved require very powerful computers indeed. According to the professor, computing power has undergone a period of enormous, exponential acceleration. He illustrates this with another comparison: Processing time has fallen so much in the last 30 years that an experiment that used to take a year can now be performed in just a second.
Nowadays, you can create a quantum mechanical description of millions of molecules on a computer and then summarize them statistically. Accordingly, this area is becoming increasingly relevant to the machine learning and artificial intelligence techniques that von Lilienfeld also uses. For example, these computer programs can recognize patterns in chemical space using statistics and make accurate quantitative predictions of the chemical behavior of new “candidate molecules.”
Understanding means predicting
“In concrete terms, we’re systematically scouring chemical space for compounds with specific properties, for example.” Other working groups then perform the corresponding experiments. “One day, our methods might actually allow people to find new medicines or valuable materials – for example, in the energy sector,” he says. Energy materials being an active field of research, the properties and a simple way to synthesize graphene, a much researched and ultrathin yet highly tear-resistant and conductive form of carbon, were discovered just a few years ago.
The professor also shows a great deal of commitment to his students: “Although not all of them will go on to become theoretical scientists, I want to stimulate their understanding of the basic principles of quantum chemistry. After all, electrons determine the behavior of molecules and can only be understood as quantum objects.” He describes what he sees as one of the fundamentals of imparting knowledge: “In chemistry, you can measure your understanding by how well you can predict molecular behavior.” That probably holds true in other disciplines as well: Those who have understood a concept correctly are able to make better predictions.
Anatole von Lilienfeld was born in the USA in 1976 and grew up in Germany. He is Associate Professor of Physical Chemistry in Basel. After completing his studies in Leipzig, Strasbourg and at ETH Zurich, he completed his doctorate at ETH Lausanne in 2005. Following research stays in the USA, he was an SNSF professor in Basel from 2013 to 2015, then an associate professor in Brussels, before returning to the University of Basel to accept a tenure track assistant professorship. Von Lilienfeld’s ancestors came from noble Russian and Baltic families that fled to Germany to escape the Russian Revolution of 1917. A father of two, he is married to a structural engineer.
More articles in the current issue of UNI NOVA.