On his career abroad and return to Russia
According to all documents I was born in Moscow, but this is not true. In fact, I was born in Dnepropetrovsk in the Ukraine, where my mother comes from. But soon after I was born we moved to Moscow. There I graduated from Moscow State University, and then went on to study at a graduate school in England, where I received a Ph.D. I worked for about a year in England. And after that I was invited to Switzerland where I became a leading researcher and head of the group. I spent six years there and defended my thesis. After that I went to the US.
I would have probably worked in the US for the rest of my days, but in 2012-2013 my life changed dramatically. My Chinese students and Russian friends persuaded me to apply for Russian and Chinese mega grants. China, too, has its own mega grants. This program is called "1000 talents". I submitted applications to both mega grants - and, to my great surprise, won both, after which I established laboratories in Russia and China. As a result, from 2013 on, three laboratories - in the US, China and Russia - started working under my supervision.
Of course, it is very difficult to live between three countries, so in the end I decided to settle in Russia, where I now spend most of my time. In 2015 I became a Professor of Skoltech, just two days ago I received the George Gamow Prize, a month ago I became a Member of the European Academy of Sciences. But I haven’t mentioned the most important thing yet: I have three children, and my fourth is on the way.
USPEX
USPEX software interface. Credit: uspex-team.org
What we are doing is a part of the new technological revolution; the name of that revolution is computational materials discovery. The dream that people have cherished for many, many decades, or even centuries is now beginning to materialize: the idea of creating materials with the power of human thought; predicting materials with new super-properties, forces; setting the stage for their experimental synthesis. And today it has already been implemented to a great extent.
A little less than two years ago a review was published that showed the pioneers of the new century of computational materials discovery. It named materials that have potential or actual application in the energy sector. At first they were predicted theoretically and only then they were synthesized experimentally and tested. These materials are for lithium-ion batteries, hydrogen storage, thermoelectric materials, photovoltaics, superconductors, materials for condensers. They were predicted by different methods. And many of them were predicted by our scientific group or by other scientific groups, using our method.
The central issue in the design of new materials is the issue of predicting their crystal structure. If you know the crystal structure, you can predict a huge list of properties using standard methods based on quantum mechanics. And even before the material is synthesized you can understand whether it will be useful.
Now, can you predict the structure of a substance that you have not yet synthesized? It is worth noting that people have learned to identify crystal structures experimentally already back in 1912. It was a breakthrough made by the Braggs father and son. In 1915, just three years after their first experiments, the Braggs received the Nobel Prize. William Lawrence Bragg, son of William Henry Bragg, who performed most of the research, received this award at 25 years of age. He is the youngest Nobel laureate in physics in its history and one of the most well-deserved. The methods for decoding the crystal structures and the lay-out that the Braggs have done are among the most important breakthroughs of the 20th century - practically everything that we know about materials is somehow connected to the knowledge of crystal structures.
William Lawrence and William Henry Braggs
The X-ray diffraction method, which the Braggs used in their research, has rendered a tremendous service to science. At least 26 Nobel prizes were given for the identification of structures of various substances by means of X-ray diffraction. Rated by the number of Nobel prizes won, crystallographers are second only to the elementary particle physicists. Many believe that one of the main scientific breakthroughs of the 20th century is the identification of the structure of DNA. But do you know how the structure of DNA was decoded? It was done by Bragg's younger colleagues in his own laboratory using X-ray diffraction and crystallography methods.
And what can we say about the theoretical prediction of structures? After all, without it, it is impossible to predict new materials with useful properties. For a long time it was believed that the prediction of structures is, in principle, an unsolvable task. For example, John Maddox wrote in his famous article: “One of the continuing scandals in the physical sciences is that it remains in general impossible to predict the structure of even the simplest crystalline solids from a knowledge of their chemical composition”.
Why are people so pessimistic? The thing is, if we want to predict the stable structure of a compound (while the stable structure is that which has minimal energy), we need to find such an arrangement of atoms in space, for which the energy will be minimal. Yes, we could go over all the possible arrangements of atoms in space, calculate the energy for each of them (which we can do: quantum mechanics gives us the necessary means for this), and then,having compared them, say which has the minimal energy.
In theory, this could be done - but not so much in practice. Different arrangements of atoms in space are astronomically many, and their number grows exponentially with the increase in the number of atoms in a recurring "box", which we call the unit cell. For example, if there are ten atoms in the unit cell, there will be about a hundred billion options for their location in space. You will need hundreds of years to go through them. And if there are 20 or 30 atoms in the unit cell, you will not have enough time, even if given the whole life of the Universe, to go through all the possible variations. And even if our computers become a trillion times more powerful, we still can not solve this problem through simple enumeration.
Hafnia crystal structure as predicted by USPEX. Credit: researchgate.net
But this does not mean that it is impossible to solve this task. That can be done in a different way. Imagine a multidimensional complex energy surface on which we need to find the lowest point. In this case, you really do not need to try every point to find what you are looking for. You just need to come up with a way out, a quick path to this global minimum. And we found this way by creating a method based on the ideas of evolution.
I named our evolutionary algorithm "Success" - USPEX, but it also has an English definition: universal structure predictor evolutionary crystallography. We published this method in 2006 and it was, indeed, very successful. Now it is being used by more than 4000 individual users, as well as several large companies such as Sony, Intel, Fujitsu, Toyota and others. Anyone can download our program from our website.
How does it work?
The strength of our method lies in the combination of a rapid evolutionary algorithm that searches for a global optimum and quantum mechanics calculations. As of now, quantum mechanics calculations are the most accurate method for calculating energies and physical properties. And the key here is the combination of speed and reliability of our algorithm with the accuracy of quantum mechanics.
Evolutionary algorithms are a special class of artificial intelligence. They are self-learning: the method can learn based on its own results and detect where the structures with low energies are located, and where the ones with high energies are. We then discard the bad structures with high energies. And we make "children” out of good structures with lower energies. So, in the next generation there are many more "children", which are in the low-energy area. Generation after generation we will test this area of low energies until we find the best structure. This way, with the help of evolution, we can increasingly focus on the most promising area of search space. And we no longer need to explore every point of our search space. We can solve this problem in a short period of time.
Confirming the unbelievable
We know perfectly well, for example, based on experiment what structure is characteristic of carbon at the high pressure of 100 GPa (this is one million atmospheres). But my program does not know anything about it. It only knows the laws of quantum mechanics and my search algorithm - the algorithm of global optimization. We make analysis that begins with random structures. They are disordered and look bad, they are not energy-efficient, but they can evolve and do it quickly enough, transforming into the correct structure. This way, without knowing anything, the method finds all that is needed.
Similarly, we have found the new allotrope of boron: a superhard substance, one of the most solid substances known to mankind at the moment. This structure was predicted and then experimentally confirmed. An even more surprising prediction, which also found its experimental confirmation, was that, at the high pressure of about two million atmospheres, sodium, according to the prediction, ceases to be a metal. We predicted that under these conditions it would become a transparent non-conductor, which actually goes against normal physical expectations. But in the course of experiments this also got confirmed.
Blind spots of chemistry
This work illustrates how we predict a stable structure for a given chemical composition, but, in fact, predictive power can be expanded massively more. We also learned how to predict which chemical compounds will be stable: without knowing the chemical formula, we can predict it. Methodologically, everything is quite simple. In thermodynamics, a stable compound is the one that has energy lower than any mixture of other compounds, presenting the same gross chemical composition. This simple criterion allows us to understand which formulae are stable, and also determine the others’ level of instability.
Here, we are in for a big surprise. For example, my Chinese Ph.D. student, who worked in Moscow and studied the manganese-boron system, managed to get some very interesting results. It’s worth noting that researchers have studied this system for many decades. He discovered all of the known stable compounds in this system with one calculation, but there were surprises along the way. And in this case we are talking about the usual conditions.
Who could have ever thought that in a system that had been studied experimentally for decades, one can get unexpected results? His calculation predicted Mn2B and MnB to have exactly the same structure as the experiments showed. But he also managed to predict a new compound, which no one has ever seen in this system - MnB3. In the course of experiments we could synthesize it and found out that it has just the same crystal structure as predicted. On the one hand, this indicates that there are still many blind spots in chemistry – and they exist even in the systems that are seemingly simple and well-researched. On the other, it clearly demonstrates that the theory has tremendous predictive power.
We also work with organic materials, particularly with polymers. We have succeeded in developing some methodological ways of predicting the structure of polymers. They were able to reproduce such structures as polyethylene and nylon. We also had a project which we carried out together with the scientists from Connecticut. On this project we were developing new polymers for flexible electronics, including condensers. We have found four new polymers that have very attractive characteristics for practical application. All four polymers were synthesized and it was confirmed that their crystal structures and dielectric properties coincided with the predictions.
Fighting complex diseases with protein prediction
By far the most important polymers for us are the proteins, which could well be referred to as the molecular machines of life. Why is it so important for us to understand their structure? Having done this, one can determine the function of a particular protein, see how it interacts with a medication, for example. But also, knowing the structure of the protein and the possible isomers of the protein structure, one can foresee whether such molecule can coagulate incorrectly.
A number of serious diseases is connected to the incorrect coagulation of protein molecules. Take, for instance, the Alzheimer's disease. In the brains of Alzheimer’s patients, the so-called amyloid plaques are formed. They prevent the spread of electrical and chemical signals. In addition, it is also believed that cancer is directly or indirectly associated with incorrect coagulation of proteins. Aging and many other diseases are also associated with it.
Can we predict the proteins? Yes, we are learning. While this work is very far from being over, we already have a powerful prototype of our method, which we have tested on very simple proteins. David Shaw - an American entrepreneur, multi-billionaire, scientist, financier, who founded his own biochemistry institute, doing its own practical research - has been doing similar work lately. His method, based on calculations that are performed using a very powerful processor, he has called Anton. In our laboratory we developed another method, based not on brute-force approach, but rather on evolutionary principles. We do not wait for the protein to coagulate anymore, as it's a very long process and we simply can not do this without a powerful computer. We force the protein to coagulate by our evolutionary method.
Their method is named Anton; we named ours Yegor (both are common Russian male names -- Ed.) Ladies and gentlemen, place your bets: will Yegor beat Anton? Initial results show that, for now, we are close to a draw: in a number of cases, David Shaw’s method has an advantage, yet, in others, ours is superior. But I think that Yegor has good prospects.
Materials of the future
We are now working with Gazprom on materials for drilling equipment. Drilling tools are based on two very good materials: diamonds, which, in the form of small crystals, are implanted in the matrix of tungsten carbide (WC). The latter is a wonderful material that has served mankind faithfully and loyally for 80 years already. In our country, based on its merit, it was named “pobedit” (Russian for "will win" -- Ed.). It was from this material that the shell caps were made for the shells during the Battle of Moscow. And these shells perforated the enemy’s tanks.
But it seems that, in the future, this material will be surpassed. We are now accumulating more and more data on materials, some of which are yet to be synthesized. However, they stand a chance of outperforming tungsten carbide both in hardness and in fracture toughness. I think that we are on the verge of a new class of materials for this kind of equipment. And we will be able to “win over pobedit”, if I’m allowed to make such a pun.
Another class of future revolutionary materials are thermoelectric materials. Here, one can come up with dozens of fantastic applications both for the direct thermoelectric effect, and for the opposite. Where can we apply thermoelectric materials? We can make refrigerators, night vision devices, devices that can detect the heat of CHP stations, thus increasing these stations’ efficiency - all based on this effect. Or we can detect and absorb the heat that is lost in your car when the engine is warming up, and turn it into electricity. Or imagine a wristwatch that is recharged with the energy of your body's heat. This technology is also available with the help of thermoelectric materials.
Use of thermoelectrical materials to produce energy. Credit: youtube.com
But there is one problem: the performance index of thermoelectric elements is still very low: it’s just a few percent or even a few fractions of a percent. According to experts, if we just double this figure, these technologies would instantly become profitable and revolutionize our lives. We are also working in this direction. We have adapted our method to search for thermoelectric materials and have tested it on the best thermoelectric material for room temperature - bismuth telluride. With the help of the calculation, we were able to find the experimentally known structure of bismuth telluride, as well as a whole family of structures that are slightly higher in energy but are many times higher in thermoelectric properties. And, based on this data, it’s realistic to increase the thermoelectric Q-factor by at least two times. But these materials still need to be created. Although, there’s nothing impossible here.
Future tasks
One can always find new tasks, and it is unlikely that we will come to a point where everything is predicted. But I think that the day will come when our calculations and research will turn from art and science that they are now into a routine. This day, on the one hand, will be very joyful for me, because this is exactly what every scientist should strive for: that their invention becomes so bug-free that everyone can use it by simply pressing the keys. My goal here is for my program to be used by all companies and scientists who develop new materials. And I want it to be as easy to use as possible.
But, on the other hand, this day will be sad, too, because when this happens, I won’t be able to do it anymore. After all, a scientist should not perform routine tasks. I will have to leave my offspring and deal with other tasks. And I hope, I will also grow fond of them as much as I did of this one.
Role models, favorite chemical elements and the importance of dreaming
Since my childhood, since I was four years old, I’d dreamed of becoming a chemist. I engrossed myself in popular science books on chemistry, the history of chemistry, books on the properties of elements. And English chemist Humphry Davy was a role model for me. He held the record of his time in the number of chemical elements he discovered. I was also impressed by the achievements of our Soviet physicists and chemists from the group of Fleurov and Oganesyan. I dreamed that, when I grow up, I would work in their laboratory and synthesize new chemical elements.
Humphry Davy
My life has worked out in such a way that my most important and unusual predictions were made using the elements that were discovered by Humphry Davy. These are sodium, magnesium and boron. And then, many years later, I happened to meet and became friends with Oganesyan. He is currently the record holder for the number of chemical elements he discovered. If I'm not mistaken, Humphry Davy had discovered 6 elements, and Oganesyan has discovered 9 elements.
So, in a slightly mystical way, my childhood dreams of following in the footsteps of Humphry Davy came true. The elements discovered by him turned out to be the most productive for my own scientific discoveries. And my dream of working in Dubna with Oganesyan and Flerov after many years transformed into my friendship with Oganesyan. So, do not be afraid to dream, because dreams really do come true.
Translated by Eugenia Romanova
