Computer synthesis of materials of a oganov. Computer Design of New Materials: Dream or Reality? You need to work with a twinkle

Artem Oganov, one of the most cited theoretical mineralogists in the world, told us about a computer prediction that has recently become achievable. Previously, this problem could not be solved because the problem of computer design of new materials includes the problem of crystal structures, which was considered unsolvable. But thanks to the efforts of Oganov and his colleagues, they managed to get closer to this dream and make it a reality.

Why this task is important: Previously, new substances were developed for a very long time and with a lot of effort.

Artem Oganov: “Experimenters go to the laboratory. Different substances are mixed at different temperatures and pressures. Get new substances. Measure their properties. As a rule, these substances are of no interest and are discarded. And experimenters are trying again to get a slightly different substance under different conditions, with a slightly different composition. And so, step by step, we overcome many failures, spending years of our lives on it. It turns out that researchers, in the hope of obtaining one material, spend a huge amount of effort, time, and also money. This process can take years. It can turn out to be a dead end and never lead to the discovery of the right material. But even when it leads to success, this success comes at a very high price. "

Therefore, it is necessary to create a technology that could make error-free predictions. That is, not to experiment in laboratories, but to instruct the computer to predict what material, with what composition and temperature, will have the desired properties under certain conditions. And the computer, going through numerous options, will be able to give an answer, what chemical composition and what crystal structure will meet the given requirements. The result may be such that the desired material does not exist. Or he is not alone.
And here a second problem arises, the solution of which is not yet available: how to get this material? That is, the chemical composition, the crystal structure is clear, but there is still no way to implement it, for example, on an industrial scale.

Prediction technology

The main thing to predict is the crystal structure. Previously, it was not possible to solve this problem, because there are many options for the arrangement of atoms in space. But the overwhelming majority of them are of no interest. What is important are those options for the arrangement of atoms in space, which are sufficiently stable and have the properties necessary for the researcher.
What are these properties: high or low hardness, electrical conductivity and thermal conductivity, and so on. The crystal structure is important.

“If you think about, say, carbon, let's look at diamond and graphite. Chemically, they are one and the same substance. But the properties are completely different. Black super soft carbon and transparent super hard diamond - what makes the difference between them? It is the crystal structure. It is thanks to her that one substance is superhard, the other is super soft. One is practically a metal conductor. The other is dielectric. "

In order to learn how to predict a new material, one must first of all learn to predict the crystal structure. For this, Oganov and his colleagues proposed an evolutionary approach in 2006.

“In this approach, we are not trying to try all the infinite number of crystal structures. We will try it out step by step, starting with a small random sample, within which we rank the possible solutions, the worst of which we discard. And from the best we produce daughter variants. Daughter variants are produced by various mutations or by recombination - by heredity, where from two parents we combine different structural features of the composition. This gives rise to a child structure - a child material, a child chemical composition, a child structure. These subsidiaries are then assessed as well. For example, by resistance or by the chemical or physical property that interests you. And those that were ranked unprofitable, we discard. Those that are promising get the right to produce offspring. We produce the next generation by mutation or heredity. "

So, step by step, scientists are approaching the material that is optimal for them in terms of a given physical property. The evolutionary approach in this case works in the same way as the Darwinian theory of evolution; this principle is implemented by Oganov and his colleagues on a computer when searching for crystal structures that are optimal from the point of view of a given property or stability.

“I can also say (but this is already a little on the verge of hooliganism) that when we were working on this method (by the way, the development continues. It has improved more and more), we experimented with different ways of evolution. For example, we tried to produce one child not from two parents, but from three or four. It turned out that, just as in life, it is optimal to produce one child from two parents. One child has two parents - dad and mom. Not three, not four, not twenty-four. This is the optimum both in nature and on a computer. "

Oganov patented his method, and now it is used by almost thousands of researchers around the world and several major companies such as Intel, Toyota and Fujitsu. Toyota, for example, according to Oganov, has for some time already, using this method, invented a new material for lithium batteries that will be used in hybrid cars.

Diamond problem

It is believed that diamond, being the record holder for hardness, is the optimal superhard material for all applications. However, this is not the case, because in iron, for example, it dissolves, and in an oxygen environment at high temperatures it burns. In general, the search for a material that would be harder than diamond has worried mankind for many decades.

“A simple computer calculation, which was carried out by my group, shows that such material cannot exist. In fact, only diamond can be harder than diamond, but in nano-crystalline form. Other materials are not able to beat a diamond in hardness. "

Another direction of Oganov's group is the prediction of new dielectric materials that could serve as the basis for super-capacitors for storing electrical energy, as well as for further miniaturization of computer microprocessors.
“This miniaturization actually meets obstacles. Because the existing dielectric materials do not withstand electrical charges quite well. They are leaking. And further miniaturization is impossible. If we can get a material that is held on silicon, but at the same time has a much higher dielectric constant than the materials we have, then we can solve this problem. And we have quite serious progress in this direction as well ”.

And the last thing that Oganov does is the development of new drugs, that is, also their prediction. This is possible due to the fact that scientists have learned to predict the structure and chemical composition of the surface of crystals.

“The fact is that the surface of a crystal often has a chemical composition that differs from the substance of the crystal itself. The structure is also very often radically different. And we found that the surfaces of simple, seemingly inert oxide crystals (such as magnesium oxide) contain very interesting ions (such as ion peroxide). They also contain groups similar to ozone, made up of three oxygen atoms. This explains one extremely interesting and important observation. When a person inhales fine particles of oxide minerals that are seemingly inert, safe and harmless, these particles play a cruel joke and contribute to the development of lung cancer. In particular, it is known that asbestos is a carcinogen, which is extremely inert. So, on the surface of such minerals as asbestos and quartz (especially quartz), peroxide ions can form, which play a key role in the formation and development of cancer. Our technique can also predict the conditions under which the formation of this kind of particles could be avoided. That is, there is even hope to find therapy and prevention of lung cancer. In this case, we are only talking about lung cancer. And from a completely unexpected side, the results of our research made it possible to understand, and maybe even prevent or cure lung cancer. "

To summarize, the prediction of crystal structures can play a key role in material design for both microelectronics and pharmaceuticals. In general, this technology opens up a new path in the technology of the future, Oganov is sure.

You can read about other areas of Artyom's laboratory by following the link, and familiarize yourself with his book Modern Methods of Crystal Structure Prediction

We are publishing the text of a lecture given by a professor at the State University of New York, an adjunct professor at Moscow State University, an honorary professor at Guilin University.Artem Oganov 8 September 2012 as part of the series "Public lectures" Polit.ru "at the open-air book festival BookMarket in the park of arts "Muzeon".

"Public lectures" Polit.ru "" are held with the support of:

Lecture text

I am very grateful to the organizers of this festival and Polit.ru for the invitation. I am honored to deliver this lecture; I hope you find it interesting.

The lecture is directly related to our future, because our future is impossible without new technologies, technologies related to our quality of life, here is the iPad, here is our projector, all our electronics, energy-saving technologies, technologies that are used to clean the environment, technologies that applied in medicine, and so on - all this depends to a large extent on new materials, new technologies require new materials, materials with unique, special properties. And there will be a story about how these new materials can be developed not in a laboratory, but on a computer.

The lecture is titled: "Computer Design of New Materials: Dream or Reality?" If this were just a dream, then the lecture would not make sense. Dreams are usually something outside the realm of reality. On the other hand, if this were already fully implemented, the lecture would also make no sense, because a new kind of methodology, including theoretical computational ones, when they are already fully developed, are moving from the category of science to the category of industrial routine tasks. In fact, this area is completely new: the computer design of new materials is somewhere in the middle between a dream - what is impossible, what we dream about at our leisure - and reality, this is an area that is not yet fully completed, this is an area that is being developed right now. And this area will allow in the near future to deviate from the traditional method of discovering new materials, laboratory, and to start computer design of materials, it would be both cheaper and faster, in many ways even more reliable. And here's how to do it, and I'll tell you. This is directly related to the problem of prediction, prediction of the structure of a substance, because the structure of a substance determines its properties. The different structure of the same substance, say, carbon, determines superhard diamond and super soft graphite. The structure in this case is everything. The structure of the substance.

In general, this year we are celebrating the centenary of the first experiments that made it possible to discover the structure of matter. A long time ago, since ancient times, people hypothesized that matter consists of atoms. A mention of this can be found, for example, in the Bible, in various Indian epics, and quite detailed references to this can be seen in Democritus and Lucretius Kara. And the first mention of how matter is arranged, how this matter consists of these discrete particles, atoms, belongs to Johannes Kepler, the great mathematician, astronomer and even astrologer - at that time astrology was still considered a science, unfortunately. Kepler drew the first pictures in which he explained the hexagonal shape of snowflakes, and the structure of ice proposed by Kepler, although different from reality, is similar in many respects to it. But, nevertheless, the hypothesis about the atomic structure of matter remained a hypothesis until the 20th century, until a hundred years ago this hypothesis became scientifically proven for the first time. It became proven with the help of my science, crystallography, a relatively new science, which was born in the middle of the 17th century, 1669 is the official date of birth of the science of crystallography, and it was created by the wonderful Danish scientist Nikolai Stenon. Actually, his name was Nils Stensen, he was a Dane, the Latinized name was Nikolai Stenon. He founded not only crystallography, but a number of scientific disciplines, and he formulated the first law of crystallography. Since that time, crystallography along an accelerating trajectory began to develop.

Nikolai Stenon had a unique biography. He became not only the founder of several sciences, but was also canonized in the Catholic Church. The greatest German poet, Goethe, was also a crystallographer. And Goethe has a quote that crystallography is unproductive, exists within itself, and in general this science is completely useless, and it is not clear why it is needed, but as a puzzle it is very interesting, and due to this it attracts very smart people. This is what Goethe said in a popular science lecture, which he gave somewhere in the spa resorts of Baden, to wealthy, lounging ladies. By the way, there is a mineral named after Goethe, goethite. I must say that at that time crystallography was really quite a useless science, really at the level of some kind of mathematical charades and puzzles. But time passed, and 100 years ago crystallography left the category of such sciences in itself and became an extremely useful science. This was preceded by a great tragedy.

I repeat, the atomic structure of matter remained a hypothesis until 1912. The great Austrian physicist Ludwig Boltzmann based all his scientific arguments on this hypothesis about the atomicity of matter and was severely criticized by many of his opponents: "How can you build all your theories on an unproven hypothesis?" Ludwig Boltzmann, influenced by this criticism and ill health, committed suicide in 1906. He hanged himself while on vacation with his family in Italy. Just 6 years later, the atomic structure of matter was proven. So if he had been a little more patient, he would have triumphed over all his opponents. Patience sometimes means more than reason, patience means more than even genius. So - what were these experiments? These experiments were carried out by Max von Laue, or rather, his graduate students. Max von Laue himself did not do any such experiments, but the idea belonged to him. The idea was that if matter really consists of atoms, if indeed, as Kepler suggested, atoms are built in a crystal in a periodic regular way, then an interesting phenomenon should be observed. X-rays were discovered not long before. Physicists by that time already well understood that if the radiation wavelength is comparable to the periodicity length - the characteristic length of an object, in this case - a crystal, then the phenomenon of diffraction should be observed. That is, the rays will travel not only strictly in a straight line, but also deviate at very strictly defined angles. Thus, some very special X-ray diffraction pattern should be observed from the crystal. It was known that the wavelength of X-rays should be similar to the size of atoms, if atoms exist, estimates of the size of the atoms were made. Thus, if the atomic hypothesis of the structure of matter is correct, then X-ray diffraction of crystals should be observed. What could be easier, how to check?

A simple idea, a simple experiment, for which in a little more than a year, Laue received the Nobel Prize in Physics. And we can try this experiment. But, unfortunately, now it is too light for everyone to observe this experiment. But maybe we can try it with one witness? Who could come up here and try to watch this experiment?

Look. Here is a laser pointer, we shine it - and what is happening here? We do not have X-rays, but an optical laser. And this is not the structure of the crystal, but its image, inflated 10 thousand times: but the laser wavelength is also 10 thousand times the wavelength of X-ray radiation, and thus the diffraction condition is fulfilled again - the comparability of the wavelength with the period of the crystal lattice. Let's take a look at an object that has no regular structure, a liquid. Here, Oleg, hold this picture, and I will shine with a laser, come closer, the picture will be small, because we cannot project ... look, you see a ring here, inside there is a point that characterizes the direct passage of the beam. But the ring is the diffraction from the unorganized structure of the liquid. If we have a crystal in front of us, then the picture will be completely different. You see, we have a lot of rays that deviate at strictly defined angles.

Oleg (volunteer): Probably because more atoms ...

Artyom Oganov: No, due to the fact that the atoms are arranged in a strictly defined way, we can observe such a diffraction pattern. This picture is very symmetrical and this is important. Let's applaud Oleg for a brilliant experiment that would have won him the Nobel Prize 100 years ago.

Then, the next year, Braggie's father and son learned to decipher diffraction pictures, to determine crystal structures from them. The first structures were very simple, but now, thanks to the latest methodologies, for which the Nobel Prize was awarded in 1985, it is possible to decipher already very, very complex structures based on experiment. Here is the experiment that Oleg and I have reproduced. Here is the initial structure, here are benzene molecules, and Oleg observed such a diffraction picture. Now, with the help of experiment, it is possible to decipher very complex structures, in particular the structures of quasicrystals, and for the discovery of quasicrystals, this new state of solid matter, last year the Nobel Prize in Chemistry was given. How dynamic is this area, what fundamental discoveries are being made in our century! The structure of proteins and other biologically active molecules is also deciphered using X-ray diffraction, this great crystallographic method.

So, we know different states of matter: ordered crystalline and quasicrystalline, amorphous (disordered solid state), as well as liquid, gaseous state and various polymeric states of matter. Knowing the structure of a substance, you can predict many, many of its properties, with a high degree of reliability. Here is the structure of magnesium silicate, a type of perovskite. Knowing the approximate positions of atoms, you can predict, for example, such a rather difficult property as elastic constants - this property is described by a rank 4 tensor with many components, and you can predict this complex property with experimental accuracy, knowing only the position of atoms. And this substance is quite important, it makes up 40% of the volume of our planet. This is the most common material on Earth. And now you can understand the properties of this substance, which exists at great depths, knowing only the arrangement of atoms.

I would like to talk a little about how properties are related to structure, how to predict the structure of a substance in order to predict new materials, and what has been done using this kind of methods. Why is ice lighter than water? We all know that icebergs float and do not sink; we know that ice is always on the surface of the river, not at the bottom. What's the matter? It's about the structure: if you look at this structure of ice, you will see large hexagonal voids in it, and when the ice begins to melt, water molecules clog these hexagonal voids, due to this, the density of the water becomes more than the density of the ice. And we can demonstrate how this process takes place. I'll show you a short film, watch it carefully. Melting will start from surfaces, which is how it actually happens, but this is a computer calculation. And you will see how the melting spreads inward ... the molecules move and you see these hexagonal channels clog up and the correctness of the structure is lost.

Ice has several different shapes, and the shape of the ice is very interesting, which is obtained by filling the voids of the ice structure with guest molecules. But the structure itself will also change. I'm talking about the so-called gas hydrates or clathrates. You see a framework of water molecules, in which there are voids in which guest molecules or atoms are present. Guest molecules can be methane - natural gas, maybe carbon dioxide, maybe, for example, a xenon atom, and each of these gas hydrates has an interesting history. The fact is that methane hydrate reserves contain 2 orders of magnitude more natural gas than traditional gas fields. Deposits of this type are located, as a rule, on the sea shelf and in permafrost zones. The problem is that humans still haven't learned how to safely and cost-effectively extract gas from them. If this problem is solved, then humanity will be able to forget about the energy crisis, we will have an almost inexhaustible source of energy for the coming centuries. Hydrate of carbon dioxide is very interesting - it can be used as a safe way of burying excess carbon dioxide. You pump carbon dioxide under slight pressure into the ice and dump it on the seabed. This ice exists there quite calmly for many thousands of years. Xenon hydrate served as an explanation for xenon anesthesia, a hypothesis that was put forward 60 years ago by the great crystal chemist Linus Pauling: the fact is that if a person is allowed to breathe xenon under slight pressure, a person ceases to feel pain. It has been used and seems to be sometimes used now for anesthesia in surgical procedures. Why?

Xenon, under low pressure, forms compounds with water molecules, forming the very gas hydrates that block the propagation of an electrical signal through the human nervous system. And the pain signal from the operated tissue simply does not reach the muscles, due to the fact that xenon hydrate is formed with just such a structure. This was the very first hypothesis, perhaps the truth is a little more complicated, but there is no doubt that the truth is near. When we talk about such porous substances, one cannot but recall microporous silicates, the so-called zeolites, which are very widely used in industry for catalysis, as well as for the separation of molecules during oil cracking. For example, octane and mesooctane molecules are perfectly separated by zeolites: this is the same chemical formula, but the structure of the molecules is slightly different: one of them is long and thin, the other is short and thick. And the one that is thin passes through the voids of the structure, and the one that is thick is eliminated, and therefore such structures, such substances are called molecular sieves. These molecular sieves are used to purify water, in particular, the water that we drink in our taps, it must go through multiple filtrations, including using zeolites. In this way, you can get rid of pollution with a wide variety of chemical contaminants. Chemical contaminants are sometimes extremely dangerous. History knows examples of how heavy metal poisoning led to very sad historical examples.

Apparently, the victims of mercury poisoning were the first first emperor of China - Qin Shi Huang Ti, and Ivan the Terrible, and the so-called mad hatter disease is very well studied, in the 18-19 centuries in England a whole class of people working in the hat industry fell ill very early a neurological disease called mad hatter disease. Their speech became incoherent, their actions senseless, their limbs trembled uncontrollably, and they fell into dementia and madness. Their body was constantly in contact with mercury, as they soaked these hats in solutions of mercury salts, which entered their bodies and affected the nervous system. Ivan the Terrible was a very progressive, good tsar under the age of 30, after that he changed overnight - and became an insane tyrant. When his body was exhumed, it turned out that his bones were severely deformed, and they contained a huge concentration of mercury. The fact is that the tsar suffered from a severe form of arthritis, and at that time arthritis was treated by rubbing in mercury ointments - this was the only remedy, and, perhaps, just mercury explains the strange madness of Ivan the Terrible. Qin Shi Huang Ti, the man who created China in its current form, ruled for 36 years, and for the first 12 years he was a puppet in the hands of his mother, the regent, his story is similar to that of Hamlet. His mother and her lover killed his father, and then they tried to get rid of him, too, a terrible story. But, having matured, he began to rule on his own - and in 12 years he stopped the internecine war between the 7 kingdoms of China, which lasted 400 years, unified China, he combined measures of weight, money, unified Chinese writing, he built the Great Wall of China, he built 6 , 5 thousand kilometers of highways that are still in use, canals that are still in use, and it was all done by one person, but in recent years he suffered from some strange form of manic insanity. His alchemists, in order to make him immortal, gave him mercury pills, they believed that this would make him immortal, as a result, this man, apparently distinguished by remarkable health, died before he was 50 years old, and the last years of this short life were clouded by madness. Lead poisoning may have made many Roman emperors victims of it: in Rome there was a lead aqueduct, an aqueduct, and it is known that with lead poisoning certain parts of the brain contract, you can even see it on tomographic pictures, intelligence falls, IQ falls, a person becomes very aggressive ... Lead poisoning is still a big problem in many cities and countries. To get rid of these kinds of unwanted effects, we need to develop new materials to clean up the environment.

Interesting materials, not fully explained, are superconductors. Superconductivity was also discovered 100 years ago. This phenomenon is in many ways exotic, it was discovered in a random way. They simply cooled the mercury in liquid helium, measured the electrical resistance, it turned out that it drops exactly to zero, and later it turned out that superconductors completely push out the magnetic field and are able to levitate in the magnetic field. These two characteristics of superconductors are widely used in high-tech applications. The type of superconductivity that was discovered 100 years ago was explained, it took half a century to explain, this explanation brought the Nobel Prize to John Bardeen and his colleagues. But then in the 80s, already in our century, a new type of superconductivity was discovered, and the best superconductors belong to this very class - high-temperature superconductors based on copper. An interesting feature is that such superconductivity still has no explanation. Superconductors have many applications. For example, with the help of superconductors, the most powerful magnetic fields are created, and this is used in magnetic resonance imaging. Magnetic hover levitating trains are another application, and here is a photo I personally took in Shanghai on such a train - you can see the speed indicator at 431 kilometers per hour. Superconductors are sometimes very exotic: organic superconductors have been known for over 30 years, that is, superconductors based on carbon, it turns out, even diamond can be made a superconductor by introducing a small amount of boron atoms into it. Graphite can also be made a superconductor.

Here is also an interesting historical parallel about how the properties of materials or their ignorance can have fatal consequences. Two stories that are very beautiful, but, apparently, are historically wrong, but I will still tell them, because a beautiful story is sometimes better than a true story. In popular science literature, one can very often find references to how the effect of the tin plague - and here is its sample - ruined the expeditions of Napoleon in Russia and Captain Scott to the South Pole. The fact is that at a temperature of 13 degrees Celsius, tin undergoes a transition from metal (this is white tin) to gray tin, a semiconductor, while the density drops sharply - and the tin falls apart. This is called the "tin plague" - the tin simply crumbles into dust. And here is a story for which I have not seen a full explanation. Napoleon comes to Russia with an army of 620 thousand, gives only a few relatively small battles - and only 150 thousand people reach Borodino. Comes 620, to Borodino almost without a fight comes 150 thousand. Under Borodino, there were about 40 thousand more victims, then the retreat from Moscow - and 5 thousand survived to Paris. By the way, the retreat was also almost without a fight. What is going on? How to slip from 620 thousand to 5 thousand without a fight? There are historians who claim that the tin plague is to blame for everything: the buttons on the uniforms of the soldiers were made of tin, the tin crumbled as soon as the cold came, and the soldiers were practically naked in the Russian frost. The problem is, the buttons were made from dirty tin, which is resistant to tin plague.

Very often you can see in the popular scientific press a mention of the fact that Captain Scott, according to various versions, either carried airplanes with him, in which fuel tanks had tin solders, or canned food in tin cans - the tin again crumbled, and the expedition died of hunger and cold. I actually read the diaries of Captain Scott - he did not mention any airplanes, he had some kind of snowmobiles, but again he does not write about the fuel tank, and he does not write about canned food either. So these hypotheses seem to be wrong, but very interesting and instructive. And remembering about the effect of the tin plague is in any case useful if you are going to a cold climate.

Here's another experience, and here I need boiling water. Another effect associated with materials and their structure, which would not have occurred to any person, is the shape memory effect, also discovered quite by accident. In this illustration, you can see that my colleagues made two letters out of this wire: TU, Technical University, they hardened this form at high temperatures. If a shape is hardened at a high temperature, the material will remember this shape. You can make a heart, for example, give a beloved one and say: this heart will remember my feelings forever ... then this shape can be destroyed, but as soon as you put it in hot water, the shape is restored, it looks like magic. You just broke this shape, but put it in hot water - the shape is restored. And all this happens due to a very interesting and rather subtle structural transformation that occurs in this material at a temperature of 60 degrees Celsius, which is why hot water is needed in our experiment. And the same transformation occurs in steel, but in steel it occurs too slowly - and the memory of the shape effect does not arise. Imagine, if steel also showed such an effect, we would live in a completely different world. The shape memory effect finds many uses: dental braces, heart shunts, engine parts in aircraft for noise reduction, adhesions in gas and oil pipelines. Now I need another volunteer ... please, what's your name? Vika? We'll need Vicky's help with this wire, it's a shape memory wire. The same alloy of nitinol, an alloy of nickel and titanium. This wire was tempered in the form of a straight wire, and he will remember this form forever. Vika, take a piece of this wire and twist it in every possible way, make it as indirect as possible, just do not tie the knots: the knot will not unravel. And now dip it in boiling water, and the wire will remember this shape ... well, how did it straighten itself? This effect can be observed forever, I have probably seen it a thousand times, but every time, like a child, I look and admire what a beautiful effect. Let's applaud Vika. It would be great if we could learn to predict such materials on a computer.

And here is the optical properties of materials, which are also completely non-trivial. It turns out that many materials, almost all crystals, split a beam of light into two beams that travel in different directions and at different speeds. As a result, if you look through the crystal at some kind of inscription, then the inscription will always be slightly doubled. But, as a rule, it is indistinguishable to our eyes. In some crystals, this effect is so strong that you can actually see two inscriptions.

Question from the audience: You said - at different speeds?

Artem Oganov: Yes, the speed of light is constant only in a vacuum. In condensed media, it is lower. Further, we used to think that each material has a specific color. Ruby is red, sapphire is blue, but it turns out that the color can also depend on the direction. In general, one of the main characteristics of a crystal is anisotropy - the dependence of properties on direction. Properties in this direction and in this direction are different. Here is the mineral cordierite, in which in different directions the color changes from brownish-yellow to blue, it is one and the same crystal. Does anyone believe me? I specially brought a crystal of cordierite, so please ... look, what color?

Question from the audience: It seems white, but like this ...

Artem Oganov: From some light, like white, to purple, you just rotate the crystal. In fact, there is an Icelandic legend about how the Vikings discovered America. And many historians see in this legend an indication of the use of this effect. When the Vikings got lost in the middle of the Atlantic Ocean, their king took out a certain sun stone, and in the twilight light managed to determine the direction to the West, and so they sailed to America. Nobody knows what a sunstone is, but many historians believe that a sunstone is what Vika holds in her hands, cordierite, by the way, cordierite is found off the coast of Norway, and with the help of this crystal you can really navigate in the twilight, in evening light, as well as in polar latitudes. And this effect was used by the US Air Force until the 50s, when it was replaced by more advanced methods. And here is another interesting effect - alexandrite, if anyone has a desire, I brought a crystal of synthetic alexandrite, and its color changes depending on the light source: daylight and electric. And finally, another interesting effect that scientists and art critics could not understand for many centuries. The Lycurgus Bowl is an object that was made by Roman artisans over 2 thousand years ago. In diffused light, this bowl is green, and in transmitted light, it is red. And I managed to understand this just a few years ago. It turned out that the bowl is not made of pure glass, but contains gold nanoparticles, which create this effect. Now we understand the nature of color - color is associated with certain absorption ranges, with the electronic structure of matter, and this, in turn, is associated with the atomic structure of matter.

Question from the audience: Can the concepts of "reflected" and "passing" be explained?

Artem Oganov: Can! By the way, I note that these very absorption spectra determine why cordierite has different colors in different directions. The fact is that the very structure of the crystal - in particular, cordierite - looks different in different directions, and light in these directions is absorbed in different ways.

What is white light? This is the entire spectrum from red to violet, and when light passes through the crystal, part of this range is absorbed. For example, a crystal can absorb blue, and what will be the result, you can see from this table. If you absorb blue rays, then the output will be orange, that is, when you see something orange, you know that this substance absorbs in the blue range. Diffused light is when you have the same bowl of Lycurgus on the table, light falls, and some of this light is scattered and gets into your eyes. Light scattering obeys completely different laws and, in particular, depends on the graininess of the object. Due to the diffusion of light, the sky is blue. There is a Rayleigh scattering law that can explain these colors.

I've shown you how properties are related to a structure. And how you can predict the crystal structure, we will look briefly now. This means that the problem of predicting crystal structures was considered unsolvable until very recently. The problem itself is formulated as follows: how to find the arrangement of atoms that gives the maximum stability - that is, the least energy? How to do it? You can, of course, enumerate all the options for the arrangement of atoms in space, but it turns out that there are so many such options that you will not have enough life to enumerate them, in fact, even for fairly simple systems, say, with 20 atoms, you will need more than time life of the universe to sort out all these possible combinations on the computer. Therefore, it was believed that this task is unsolvable. Nevertheless, it was possible to solve this problem, moreover by several methods, and the most effective method, although it may sound immodest, was developed by my group. The method is called "Success", "USPEX", an evolutionary method, an evolutionary algorithm, the essence of which I will try to explain to you now. The task is equivalent to finding the global maximum on some multidimensional surface - for simplicity, consider a two-dimensional surface, the surface of the Earth, where you need to find the highest mountain without having maps. Let's put it this way, as my Australian colleague Richard Clegg put it - he is Australian, he loves kangaroos, and in his formulation with the help of kangaroos, enough non-intelligent animals, you need to determine the highest point on the surface of the Earth. The kangaroo only understands simple instructions - go up, go down. In the evolutionary algorithm, we throw off a landing of kangaroos, randomly, to different parts of the planet and give each of them an instruction: go up to the top of the nearest hill. And they go. When these kangaroos reach Sparrow Hills, for example, and when they reach Elbrus, it may be, and those of them who have not reached high are eliminated and shot back. A hunter comes, he almost said an artist, a hunter comes and shoots, and those who survived get the right to breed. And thanks to this, it is possible to select the most promising areas from the entire search space. And step by step, by shooting taller and taller kangaroos, you will move the kangaroo population to the global maximum. Kangaroos will produce more and more successful offspring, hunters will shoot more and more high climbing kangaroos, and thus this population can simply be driven to Everest.

And this is the essence of evolutionary methods. For the sake of simplicity, I omit the technical details of exactly how this was implemented. And here is another two-dimensional implementation of this method, here is the surface of energies, we need to find the bluest point, here are our initial, random, structures - these are the bold points. The calculation immediately understands which of them are bad, here - in the red and yellow areas, which of them are the most promising: in the blue, greenish areas. And step by step the density of testing the most promising areas grows until we find the most suitable, most stable structure. There are different methods for predicting structures - methods of random search, artificial annealing, and so on, but the most powerful method turned out to be this evolutionary one.

The most difficult thing is how to produce offspring from parents on a computer. How to take two parent structures and make a child out of them? In fact, on a computer, you can make children not only from two parents, we experimented, we out of three, and out of four we tried to do it. But, as it turns out, this does not lead to anything good, just like in life. A child is better off having two parents. By the way, one parent also works, two parents are optimal, and three or four are no longer working. The evolutionary method has several interesting features that, by the way, make it akin to biological evolution. We see how from the unadapted, random structures from which we start the calculation, in the course of the calculation, highly organized, highly ordered solutions appear. We see that the calculations are most effective when the population of structures is the most diverse. The most stable and most surviving populations are populations of diversity. For example, what I like about Russia is that there are more than 150 peoples in Russia. There are fair-haired, there are dark-haired, there are all sorts of people of Caucasian nationality like me, and all this gives the Russian population stability and future. Monotonous populations have no future. This can be seen from the efficiency calculations very clearly.

Can we predict that graphite is the stable form of carbon at atmospheric pressures? Yes. This calculation is very fast. But in addition to graphite, we produce several interesting slightly less stable solutions in the same calculation. And these solutions can be interesting too. If we increase the pressure, the graphite is already unstable. A diamond is stable, and we find it very easily too. See how the calculation quickly produces a diamond from disordered initial structures. But before a diamond is found, a number of interesting structures are produced. For example, this structure. While the diamond has hexagonal rings, the 5 and 7-sided rings are visible here. This structure is only slightly inferior in stability to diamond, and at first we thought it was a curiosity, but then it turned out that this is a new, really existing form of carbon, which was recently established by us and our colleagues. This calculation was made at 1 million atmospheres. If we increase the pressure to 20 million atmospheres, the diamond will cease to be stable. And instead of diamond, a very strange structure will be stable, the stability of which for carbon at such pressures has been guessed for many decades, and our calculation confirms this.

A lot has been done by both us and our colleagues using this method, here is a small selection of different discoveries. Let me tell you just a few of them.

Using this method, you can replace the laboratory discovery of materials with a computer one. In laboratory material discovery, Edison was the unsurpassed champion, who said, "I didn't fail 10,000, I just found 10,000 ways that don't work." This tells you how many attempts it takes, unsuccessful attempts to make before making a real discovery by this method, and with the help of computer design you can achieve success in 1 attempt out of 1, in 100 out of 100, in 10 thousand out of 10 thousand, this is our the goal is to replace Edison's method with something much more productive.

We can now optimize not only energy, but also any property. The simplest property is density, and diamond is the densest material known so far. Almaz is generally a record holder in many ways. A cubic centimeter of diamond contains more atoms than a cubic centimeter of any other substance. Diamond holds the record for hardness, and it is also the least compressible substance known. Can these records be broken? Now we can ask this question to the computer, and the computer will give the answer. And the answer is yes, some of these records can be broken. It turned out that the density of a diamond is easy enough to beat, there are denser forms of carbon that have a right to exist, but have not yet been synthesized. These forms of carbon beat diamond not only in density but also in optical properties. They will have higher refractive indices and light dispersion - what does this mean? The refractive index of a diamond gives a diamond its unrivaled brilliance and internal light reflection - and light dispersion means that white light will split into a red to violet spectrum even more than a diamond does. By the way, here is the material that often replaces diamond in the jewelry industry - cubic zirconia, cubic zirconia. It surpasses diamond in light dispersion, but, unfortunately, is inferior to diamond in brilliance. And new forms of carbon will beat diamond on both counts. What about hardness? Until 2003, it was believed that hardness is a property that people will never learn to predict and calculate, in 2003 everything changed with the work of Chinese scientists, and this summer I visited Yangshan University in China, where I received another honorary professor degree, and there I visited the founder of this whole theory. We managed to develop this theory.

Here is a table that shows how the calculated hardness determinations agree with experiment. For most normal substances, the agreement is excellent, but for graphite, the models predicted that it should be superhard, which is obviously wrong. We managed to understand and fix this error. And now, using this model, we reliably predict the hardness for any substance, and we can ask the computer the following question: which substance is the hardest? Can a diamond be surpassed in hardness? People have actually thought about this for many, many decades. So what's the hardest structure for carbon? The answer was discouraging: diamond, and nothing harder in carbon can be. But you can find structures of carbon that will be close in hardness to diamond. Carbon structures that are close to diamond in hardness do indeed have a right to exist. And one of them is the one I showed you earlier, with 5- and 7-member channels. Dubrovinsky in 2001, an ultra-solid substance, titanium dioxide, was proposed in the literature; it was believed that in terms of hardness it was not much inferior to diamond, but there were doubts. The experiment was controversial enough. Almost all experimental measurements from that work were sooner or later refuted: it was very difficult to measure the hardness, due to the small size of the samples. But the calculation showed that the hardness was also erroneously measured in that experiment, and the real hardness of titanium dioxide is about 3 times less than what the experimenters claimed. So with the help of this kind of calculations, one can even judge which experiment is reliable and which one is not, so much these calculations have now reached high accuracy.

There is another story related to carbon that I would like to tell you - it has been especially violent in the last 6 years. But it began 50 years ago, when American researchers conducted such an experiment: they took graphite and compressed it to a pressure of about 150-200 thousand atmospheres. If graphite is compressed at high temperatures, it must transform into diamond, the most stable form of carbon at high pressures - this is how diamond is synthesized. If you do this experiment at room temperature, then the diamond cannot be formed. Why? Because the rearrangement of the structure that is required to transform graphite into diamond is too large, these structures are too dissimilar, and the energy barrier to be overcome is too large. And instead of diamond formation, we will observe the formation of some other structure, not the most stable one, but the one with the lowest formation barrier. We proposed such a structure - and called it M-carbon, this is the same structure with 5- and 7-membered rings; my Armenian friends jokingly call him "Muglerod-Shmuglerod". It turned out that this structure fully describes the results of that 50-year-old experiment, and the experiment was repeated many times. The experience, by the way, is very beautiful - by compressing graphite (a black, soft opaque semi-metal) at room temperature, under pressure, the researchers obtained a transparent superhard non-metal: an absolutely fantastic transformation! But this is not a diamond, its properties are not consistent with diamond, and our hypothetical structure then fully described the properties of this substance. We were terribly happy, wrote an article and published it in the prestigious journal Physical Review Letters, and rested on our laurels for exactly one year. A year later, American and Japanese scientists found a new structure, completely different from it, this one, with 4- and 8-membered rings. This structure is completely different from ours, but describes the experimental data almost as well. The problem is that the experimental data were of low resolution, and many other structures were suitable for them. Another six months passed, a Chinese named Wang suggested W-carbon, and W-carbon also explained the experimental data. Soon the story became grotesque - new Chinese groups joined it, and the Chinese love to produce, and they stamped about 40 structures, and they all fit the experimental data: P-, Q-, R-, S-carbon, Q-carbon, X -, Y-, Z-carbon, M10-carbon is known, X'-carbon, and so on - the alphabet is already lacking. So who's right? Generally speaking, our M-carbon initially had exactly the same amount of claims to be right as everyone else.

A remark from the audience: Everyone is right.

Artem Oganov: This also does not happen! The fact is that nature always chooses extreme solutions. Not only people are extremists, but nature is also extremist. At high temperatures, nature chooses the most stable state, because at high temperatures, you can go through any energy barrier, and at low temperatures, nature chooses the smallest barrier, and there can be only one winner. There can only be one champion - but who exactly? You can do a high resolution experiment, but people tried for 50 years and no one succeeded, all the results were of poor quality. You can make a calculation. And in the calculation, one could consider the activation barriers to the formation of all these 40 structures. But, firstly, the Chinese are still stamping new and new structures, and no matter how much you try, there will still be some Chinese who will say: but I have one more structure, and you will count these activation barriers until you are sent to a well-deserved rest. This is the first difficulty. The second difficulty is that it is very, very difficult to count activation barriers in solid-state transformations, this is an extremely nontrivial task, special methods and powerful computers are needed. The fact is that these transformations do not take place in the entire crystal, but first in a small fragment - the embryo, and then spreads into the embryo and further. And modeling this embryo is an extremely difficult task. But we found such a method, a method that had been developed earlier by Austrian and American scientists, and we adapted it to our task. We managed to modify this method so that with one blow we were able to solve this problem once and for all. We set the problem as follows: if you start with graphite, the initial state is rigidly specified, and the final state is given vaguely - any tetrahedral, sp3-hybridized form of carbon (and these are the states we expect under pressure), which of the barriers will be the minimum? This method is able to count barriers and finds the minimum barrier, but if we set the final state as an ensemble of different structures, then we can solve the problem completely. We started the calculation with the path of the graphite - diamond transformation as a "seed", we know that this transformation is not observed in the experiment, but we were interested in what the calculation would do with this transformation. We waited a little (in fact, this calculation took six months on a supercomputer) - and instead of a diamond, the calculation gave us M-carbon.

In general, I must say, I am an extremely lucky person, I had 1/40 chances of winning, because there were about 40 structures that had equal chances to win, but I pulled out the lottery ticket again. Our M-carbon won, we published our results in the prestigious new journal Scientific Reports, the new journal of the Nature group, and a month after we published our theoretical results, the same journal published the results of a high-resolution experiment, for the first time in 50 years. received. Researchers at Yale University did a high-resolution experiment and tested all these structures, and it turned out that only M-carbon fits all the experimental data. And now in the list of forms of carbon there is another experimentally and theoretically established allotrope of carbon, M-carbon.

I will mention one more alchemical transformation. Under pressure, it is expected that all substances will turn into a metal, sooner or later any substance will become a metal. And what will happen to the substance, which is already metal? For example sodium. Sodium is not just a metal at all, but an amazing metal described by the free electron model, that is, it is the limiting case of a good metal. What happens if sodium is squeezed? It turns out that sodium will cease to be a good metal - at first, sodium will turn into a one-dimensional metal, that is, electricity will only conduct in one direction. At higher pressures, we predicted that sodium would lose its metallicity altogether and turn into a reddish transparent dielectric, and if the pressure was increased even more, it would become colorless, like a glass. So - you take a silvery metal, squeeze it - at first it turns into a bad metal, black as coal, you squeeze further - it turns into a reddish transparent crystal that looks like a ruby, and then it becomes white like a glass. We predicted it, and the journal Nature, where we sent it, refused to publish it. The editor returned the text within a few days and said: we don't believe, it's too exotic. We found an experimenter, Mikhail Eremts, who was ready to test this prediction - and here is the result. At 110 Gigapascal, that's 1.1 million atmospheres, it's still a silvery metal, at 1.5 million atmospheres, it's a bad metal, black as coal. At 2 million atmospheres, it is a transparent, reddish non-metal. And already with this experiment, we very easily published our results. This, by the way, is a rather exotic state of matter, because the electrons are no longer smeared in space (as in metals) and are not localized on atoms or bonds (as in ionic and covalent substances) - valence electrons, which ensured metallicity to sodium, are trapped in voids space, where there are no atoms, and they are very localized. Such a substance can be called an electride, i.e. salt, where the role of negatively charged ions, anions, is played not by atoms (say, fluorine, chlorine, oxygen), but by bunches of electron density, and our sodium form is the simplest and most striking example of an electride known.

You can use this kind of calculations and to understand the substance of the earth's and planetary interior. We learn about the state of the earth's interior mainly from indirect data, from seismological data. We know that there is a metallic core, mainly consisting of iron, the Earth's core, and a non-metallic shell, consisting of magnesium silicates, called the mantle, and at the very surface there is a thin earth crust on which we live, and which we know very much OK. And the interior of the Earth is almost completely unknown to us. By direct testing, we can study only the very surface of the Earth. The deepest well is the Kola superdeep, its depth is 12.3 kilometers, drilled in the USSR, no one could drill further. The Americans tried to drill, went broke on this project and stopped it. They invested huge sums in the USSR, drilled up to 12 kilometers, then perestroika happened, and the project was frozen. But the radius of the Earth is 500 times larger, and even the Kola superdeep borehole drilled only the very surface of the planet. But the substance of the depths of the Earth determines the face of the Earth: earthquakes, volcanism, continental drift. A magnetic field forms in the core of the Earth, which we will never reach. Convection of the molten outer core of the Earth is responsible for the formation of the Earth's magnetic field. By the way, the inner core of the Earth is solid, and the outer is molten, it is like chocolate candy with molten chocolate, and inside is a nut - this is how you can imagine the core of the Earth. Convection of the solid mantle of the Earth is very slow, its speed is on the order of 1 centimeter per year; hotter currents go up, colder ones go down, and this is the convective movement of the Earth's mantle and is responsible for continental drift, volcanism, earthquakes.

An important question is what is the temperature at the center of the Earth? We know the pressure from seismological models, but these models do not give temperature. Temperature is defined as follows: we know that the inner core is solid, the outer core is liquid, and that the core is made of iron. Thus, if you know the melting point of iron at that depth, then you know the temperature of the core at that depth. Experiments were made, but they gave an uncertainty of 2 thousand degrees, and calculations were made, and the calculations put an end to this question. The melting point of iron at the border of the inner and outer core was about 6.4 thousand Kelvin. But when geophysicists learned about this result, it turned out that this temperature is too high in order to correctly reproduce the characteristics of the Earth's magnetic field - this temperature is too high. And then physicists remembered that, in fact, the nucleus is not pure iron, but contains various impurities. We still don't know exactly which ones, but among the candidates are oxygen, silicon, sulfur, carbon, hydrogen. By varying different impurities, comparing their effects, it was possible to understand that the melting temperature should be lowered by about 800 degrees. 5600 degrees Kelvin is such a temperature at the border of the inner and outer cores of the Earth, and this estimate is currently generally accepted. This effect of lowering the temperature of impurities, the eutectic lowering of the melting point, is well known, thanks to this effect, our shoes suffer in winter - the roads are sprinkled with salt in order to lower the melting point of the snow, and thanks to this solid snow, the ice turns into a liquid state, and our shoes suffer from of this salt water.

But perhaps the most powerful example of the same phenomenon is Wood's alloy - an alloy that consists of four metals, there are bismuth, lead, tin and cadmium, each of these metals has a relatively high melting point, but the effect of mutual lowering of the melting point works so much that Wood's alloy melts in boiling water. Who wants to do this experience? By the way, I bought this sample of Wood's alloy in Yerevan on the black market, which will probably give this experience an additional flavor.

Pour boiling water while I hold Wood's alloy and you will see the drops of Wood's alloy falling into the glass.

Drops fall - that's enough. It melts at hot water temperature.

And this effect occurs in the core of the Earth, due to this, the melting point of the ferrous alloy decreases. But now the next question is: what does the core consist of? We know that there is a lot of iron and there are some light impurity elements, we have 5 candidates. We started with the least likely candidates — carbon and hydrogen. I must say that until recently, few people paid attention to these candidates, both were considered unlikely. We decided to check it out. With an employee of Moscow State University Zulfiya Bazhanova, we decided to tackle this business, predict stable structures and stable compositions of iron carbides and hydrides in the conditions of the Earth's core. We also did this for silicon, where we did not find any special surprises - and for carbon, it turned out that those compounds that were considered stable for many decades, in fact, at the pressures of the Earth's core turn out to be unstable. And it turns out that carbon is a very good candidate, in fact, carbon alone can explain many of the properties of the Earth's inner core ideally, contrary to previous work. Hydrogen turned out to be a rather bad candidate, hydrogen alone cannot explain any properties of the Earth's core. Hydrogen can be present in small quantities, but it cannot be the main impurity element in the Earth's core. For hydrogen hydrides under pressure, we found a surprise - it turned out that there is a stable compound with a formula that contradicts school chemistry. A normal chemist will write the formula of hydrogen hydrides as FeH 2 and FeH 3, generally speaking, FeH also appears under pressure, and they put up with this - but the fact that FeH 4 can appear under pressure was a real surprise. If our kids in school write the formula FeH 4, I guarantee they will get a grade in chemistry, most likely even in the quarter. But it turns out that under pressure, the rules of chemistry are violated - and such exotic compounds appear. But, as I said, iron hydrides are unlikely to be important for the interior of the Earth, it is unlikely that hydrogen is present there in significant quantities, but carbon is most likely present.

And finally, the last illustration, about the Earth's mantle, or rather, about the boundary between the core and the mantle, the so-called "D" layer, which has very strange properties. One of the properties was the anisotropy of propagation of seismic waves, sound waves: in the vertical direction and in the horizontal direction, the velocities differ significantly. Why is this so? For a long time I could not understand. It turns out that a new structure of magnesium silicate is formed in the layer at the boundary between the core and the mantle of the Earth. We managed to understand this 8 years ago. At the same time, we and our Japanese colleagues published 2 papers in Science and Nature, which proved the existence of this new structure. It is immediately evident that this structure looks completely different in different directions, and its properties should differ in different directions - including the elastic properties, which are responsible for the propagation of sound waves. With the help of this structure, it was possible to explain all those physical anomalies that were discovered and caused trouble for many, many years. They even managed to make several predictions.

In particular, on smaller planets such as Mercury and Mars, there will be no layer like D ”. There will not be enough pressure to stabilize this structure. It was also possible to make a prediction that as the Earth cools, this layer should grow, because the stability of post-perovskite increases with decreasing temperature. It is possible that when the Earth was formed, this layer did not exist at all, but it was born in the early phase of the development of our planet. And now all this can be understood thanks to the predictions of new structures of crystalline substances.

A remark from the audience: Thanks to a genetic algorithm.

Artem Oganov: Yes, although this last story about post-perovskite preceded the invention of this evolutionary method. By the way, she prompted me to invent this method.

A remark from the audience: So 100 years of this genetic algorithm, there that just did not do it.

Artem Oganov: This algorithm was created by me and my graduate student in 2006. By the way, it is wrong to call it "genetic"; a more correct name is "evlution". Evolutionary algorithms appeared in the 70s, and they have found application in very many areas of technology and science. For example, cars, ships and airplanes are optimized using evolutionary algorithms. But for each new task, the evolutionary algorithm is completely different. Evolutionary algorithms are not one method, but a huge group of methods, a whole huge area of ​​applied mathematics, and for each new type of problem a new approach must be invented.

A remark from the audience: What is the math? Genetics is.

Artem Oganov: This is not genetics - this is mathematics. And for each new task, you need to invent your new algorithm from scratch. And humans have actually tried before us to invent evolutionary algorithms and adapt them to predict crystal structures. But they took algorithms from other areas too literally - and it didn't work, so we had to create a new method from scratch, and it turned out to be very powerful. Although the field of evolutionary algorithms has been around as long as I have - since at least 1975, it took a lot of effort to predict crystal structures to create a working method.

All these examples that I have given you show how understanding the structure of matter and the ability to predict the structure of matter lead to the design of new materials that can have interesting optical properties, mechanical properties, and electronic properties. Materials that make up the bowels of the Earth and other planets. In this case, you can solve a whole range of interesting problems on a computer using these methods. A huge contribution to the development of this method and its application was made by my colleagues and more than 1000 users of our method in different parts of the world. Let me sincerely thank all these people and the organizers of this lecture, and you - for your attention.

Lecture discussion

Boris Dolgin: Thanks a lot! Thank you very much, Artyom, thank you very much to the organizers who gave us a platform for this version of public lectures, thank you very much to RVC, which supported us in this initiative, I am sure that Artyom's research will continue, which means that new material will appear for his lecture with us, here , because I must say that some of what was heard today actually did not exist at the time of the previous lectures, so it makes sense.

Question from the audience: Please tell me how to ensure the room temperature at such a high pressure? Any system of plastic deformation is accompanied by heat release. Unfortunately, you did not mention this.

Artem Oganov: The point is, it all depends on how fast you compress. If compression is carried out very quickly, for example, in shock waves, then it is necessarily accompanied by heating, a sharp compression necessarily leads to an increase in temperatures. If you compress it slowly, then the sample has enough time to exchange heat with its environment and come into thermal equilibrium with its environment.

Question from the audience: And did your installation allow you to do this?

Artem Oganov: The experiment was not carried out by me, I did only calculations and theory. I do not admit myself to the experiment due to internal censorship. And the experiment was carried out in chambers with diamond anvils, where a sample is squeezed between two small diamonds. In such experiments, the sample has so much time to come to thermal equilibrium that the question does not arise here.

The essence of the search for the most stable structure is reduced to calculating the state of matter that has the lowest energy. The energy in this case depends on the electromagnetic interaction of the nuclei and electrons of the atoms that make up the crystal under study. It can be estimated using quantum mechanical calculations based on the simplified Schrödinger equation. So the USPEX algorithm uses density functional theory, which developed in the second half of the last century. Its main purpose is to simplify calculations of the electronic structure of molecules and crystals. The theory allows replacing the many-electron wave function with the electron density, while remaining formally accurate (but in fact the approximations turn out to be inevitable). In practice, this leads to a decrease in the complexity of calculations and, as a consequence, the time that will be spent on them. Thus, quantum mechanical calculations are combined with the evolutionary algorithm in USPEX (Fig. 2). How does the evolutionary algorithm work?

Search for structures with the lowest energy can be enumerated: randomly arrange atoms relative to each other and analyze each such state. But since the number of options is huge (even if there are only 10 atoms, then the possibilities of their location relative to each other will be about 100 billion), the calculation would take too long. Therefore, scientists managed to achieve success only after developing a more cunning method. The USPEX algorithm is based on an evolutionary approach (Figure 2). First, a small number of structures are randomly generated and their energy is calculated. The system removes the variants with the highest energy, that is, the least stable ones, and generates similar ones from the most stable ones and calculates them already. Simultaneously, the computer continues to randomly generate new structures to maintain the diversity of the population, which is an essential condition for successful evolution.

Thus, logic taken from biology helped to solve the problem of predicting crystal structures. It is difficult to say that this system contains a gene, because new structures can differ from their predecessors in very different parameters. The most adapted to the conditions of selection "individuals" leave offspring, that is, the algorithm, learning from its mistakes, maximizes the chances of success in the next attempt. The system quickly finds the option with the lowest energy and efficiently calculates the situation when the structural unit (cell) contains tens and even the first hundreds of atoms, while the previous algorithms could not cope with ten.

One of the new challenges facing USPEX at Moscow Institute of Physics and Technology is to predict the tertiary structure of proteins from their amino acid sequence. This problem of modern molecular biology is one of the key ones. In general, the task before scientists is very difficult also because it is difficult to calculate the energy for such a complex molecule as protein. According to Artem Oganov, his algorithm already manages to predict the structure of peptides with a length of about 40 amino acids.

Video 2. Polymers and biopolymers. What substances are polymers? What is the structure of the polymer? How common is the use of polymeric materials? Professor, PhD in Crystallography Artem Oganov talks about this.

Explanation of USPEX

In one of his popular science articles, Artem Oganov (Fig. 3) describes USPEX as follows:

“Here is a vivid example to demonstrate the general idea. Imagine that you need to find the highest mountain on the surface of an unknown planet, on which complete darkness reigns. In order to save resources, it is important to understand that we do not need a complete relief map, but only its highest point.

Figure 3. Artem Romaevich Oganov

You land a small assault force of biorobots on the planet, sending them one by one to random places. The instruction that each robot must follow is to walk along the surface against the forces of gravitational attraction and eventually reach the top of the nearest hill, the coordinates of which it must report to the orbital base. We do not have the funds for a large research contingent, and the likelihood that one of the robots will immediately climb the highest mountain is extremely small. This means that it is necessary to apply the well-known principle of Russian military science: “fight not by number, but by skill,” which is realized here in the form of an evolutionary approach. Bearing the nearest neighbor, robots meet and reproduce their own kind, placing them along the line between “their” peaks. The progeny of biorobots proceeds to carry out the same instructions: they move in the direction of the elevation of the relief, exploring the area between the two peaks of their “parents”. Those “individuals” that have found peaks below the average level are recalled (this is how selection is carried out) and re-dropped randomly (this is how the maintenance of the “genetic diversity” of the population is modeled). "

How to estimate the error with which USPEX works? You can take a problem with a known correct answer and independently solve it 100 times using an algorithm. If the correct answer is obtained in 99 cases, then the probability of a calculation error will be 1%. Usually correct predictions are obtained with a probability of 98–99% when the number of atoms in a unit cell is 40.

The evolutionary USPEX algorithm has led to many interesting discoveries and even the development of a new dosage form of a medicine, which will be discussed below. I wonder what will happen when new generation supercomputers appear? Will the crystal structure prediction algorithm fundamentally change? For example, some scientists are developing quantum computers. In the future, they will be much more effective than the most advanced modern ones. According to Artem Oganov, evolutionary algorithms will retain their leading position, but they will start to work faster.

Areas of laboratory work: from thermoelectrics to drugs

USPEX turned out to be an algorithm that is not only effective, but also multifunctional. At the moment, under the leadership of Artem Oganov, a lot of scientific works in various directions are being carried out. Some of the latest projects are attempts to model new thermoelectric materials and predict the structure of proteins.

“We have several projects, one of them is the study of low-dimensional materials such as nanoparticles, material surfaces, Another is the study of chemicals under high pressure. There is also an interesting project related to the prediction of new thermoelectric materials. Now we already know that the adaptation of the method for predicting crystal structures, which we have invented, to the problems of thermoelectricity works effectively. At the moment, we are already ready for a big leap forward, the result of which should be the discovery of new thermoelectric materials. It is already clear that the method that we have created for thermoelectrics is very powerful, the tests carried out are successful. And we are fully prepared to look for new materials proper. We are also engaged in the prediction and study of new high-temperature superconductors. We ask the question of predicting the structure of proteins. This is a new task for us and a very interesting one ”.

Interestingly, USPEX has already benefited even medicine: “Moreover, we are developing new medicines. In particular, we predicted, obtained and patented a new drug,- says A.R. Oganov. - This is 4-aminopyridine hydrate, a drug for multiple sclerosis. ".

We are talking about a drug recently patented by the staff of the laboratory of computer-aided design of materials Valery Royzen (Fig. 4), Anastasia Naumova and Artem Oganov, which allows symptomatic treatment of multiple sclerosis. The patent is open, which will help reduce the price of the drug. Multiple sclerosis is a chronic autoimmune disease, that is, one of those pathologies when the host's own immune system harms the host. In this case, the myelin sheath of nerve fibers is damaged, which normally performs an electrically insulating function. It is very important for the normal functioning of neurons: current through the outgrowths of nerve cells covered with myelin is carried out 5-10 times faster than through uncovered ones. Therefore, multiple sclerosis leads to disturbances in the functioning of the nervous system.

The underlying causes of multiple sclerosis remain unclear. Many laboratories around the world are trying to understand them. In Russia, this is done by the laboratory of biocatalysis at the Institute of Bioorganic Chemistry.

Figure 4. Valery Royzen - one of the authors of the patent for a drug for multiple sclerosis, an employee of the laboratory of computer-aided design of materials, developing new dosage forms of medicines and actively promoting science.

Video 3. Popular science lecture by Valery Roysen "Tasty Crystals". You will learn about the principles of how drugs work, the importance of the form of drug delivery to the human body and about the evil twin brother of aspirin.

Previously, 4-aminopyridine was already used in the clinic, but scientists managed, by changing the chemical composition, to improve the absorption of this drug into the blood. They obtained crystalline 4-aminopyridine hydrate (Fig. 5) with a stoichiometry of 1: 5. In this form, the medicine itself and the method of its preparation were patented. The substance improves the release of neurotransmitters in the neuromuscular synapses, which makes it easier for patients with multiple sclerosis. It is worth noting that this mechanism involves the treatment of symptoms, but not the disease itself. In addition to bioavailability, a fundamental point in the new development is the following: since it was possible to "enclose" 4-aminopyridine in a crystal, it became more convenient for use in medicine. Crystalline substances are relatively easy to obtain in a purified and homogeneous form, and the freedom of the drug from potentially harmful impurities is one of the key criteria for a good medicine.

Discovery of new chemical structures

As mentioned above, USPEX allows you to find new chemical structures. It turns out that even the "usual" carbon has its own mysteries. Carbon is a very interesting chemical element because it forms a wide range of structures, from superhard dielectrics to soft semiconductors and even superconductors. The former include diamond and lonsdaleite, the latter - graphite, and the third - some fullerenes at low temperatures. Despite the wide variety of known forms of carbon, scientists led by Artem Oganov managed to discover a fundamentally new structure: it was not previously known that carbon can form guest-host complexes (Fig. 6). The work was attended, among other things, by the staff of the laboratory of computer-aided design of materials (Fig. 7).

Figure 7. Oleg Feya, post-graduate student at Moscow Institute of Physics and Technology, employee of the laboratory of computer-aided design of materials and one of the authors of the discovery of a new structure of carbon. In his free time, Oleg is engaged in the popularization of science: his articles can be read in the publications "Schrödinger's Cat", "For Science", STRF.ru, "Strana Rosatom". In addition, Oleg is the winner of the Moscow Science Slam and a participant in the TV show "The Smartest".

The “guest-host” interaction is manifested, for example, in complexes consisting of molecules that are connected to each other by non-covalent bonds. That is, a certain atom / molecule occupies a certain place in the crystal lattice, but at the same time does not form a covalent bond with the surrounding compounds. This behavior is widespread among biological molecules that bind to each other, forming strong and large complexes that perform various functions in our body. In general, we mean compounds consisting of two types of structural elements. For substances formed only by carbon, such forms were not known. Scientists published their discovery in 2014, expanding our knowledge of the properties and behavior of the 14th group of chemical elements as a whole (Fig. 8). It should be noted that in the open form of carbon, covalent bonds between atoms are formed. We are talking about the type of guest-host because of the presence of two distinct types of carbon atoms, which have completely different structural environments.

New chemistry under high pressure

In the laboratory of computer-aided design of materials, they study which substances will be stable at high pressures. Here is how the head of the laboratory argues the interest in such research: “We study materials under high pressure, in particular the new chemistry that emerges under these conditions. This is a very unusual chemistry that does not fit into the rules of the traditional one. The knowledge gained about new compounds will lead to an understanding of what is happening inside the planets. Because these unusual chemicals can prove to be very important materials in the planet's interior. " It is difficult to predict how substances under high pressure will behave: most chemical rules stop working because these conditions are very different from what we are used to. Nevertheless, it is necessary to understand this if we want to know how the Universe works. The lion's share of the baryonic matter of the Universe is under high pressure inside planets, stars, satellites. Surprisingly, very little is known about its chemistry.

New chemistry, which is realized at high pressure in the laboratory of computer-aided design of materials of the MIPT, is being studied by PhD (degree similar to Ph.D.) Gabriele Saleh:

“I am a chemist and I am interested in chemistry at high pressures. Why? Because we have the rules of chemistry that were formulated 100 years ago, but recently it turned out that they stop working at high pressures. And this is very interesting! It looks like an amusement park: there is a phenomenon that no one can explain; exploring a new phenomenon and trying to understand why it happens is very interesting. We started our conversation with fundamental things. But high pressures also exist in the real world. Of course, not in this room, but inside the Earth and on other planets " .

Since I’m a chemist I’m interested in high-pressure chemistry. Why? Because we have chemical rules which were established one hundred years ago but recently it was discovered that these rules get broken at high pressure. And it is very interesting! This is like a loonopark because you have a phenomenon, which nobody can rationalize. It's interesting to study new phenomenon and to try to understand why does it happen. We started from the fundamental point of view. But these high pressures exist. Not in this room of course but in the inside of the Earth and in other planets.

Figure 9. Carbonic acid (H 2 CO 3) - structure stable under pressure. In the insert above it is shown that along C axis polymer structures are formed. Studying the carbon-oxygen-hydrogen system under high pressures is very important for understanding how the planets work. H 2 O (water) and CH 4 (methane) are the main constituents of some giant planets, such as Neptune and Uranus, where pressures can reach hundreds of GPa. Large ice satellites (Ganymede, Callisto, Titan) and comets also contain water, methane and carbon dioxide, which are subject to pressures of up to several GPa.

Gabriele told us about his new work, which has recently been accepted for publication:

“Sometimes you do basic science, but then you find a direct application of the knowledge gained. For example, we recently submitted an article for publication describing the search results for all stable compounds made from carbon, hydrogen and oxygen at high pressure. We found one that is stable at very low pressures such as 1 GPa , and it turned out to be carbonic acid H 2 CO 3(fig. 9). I studied the literature on astrophysics and found that the moons Ganymede and Callisto [moons of Jupiter] are composed of water and carbon dioxide: molecules that form carbonic acid. Thus, we realized that our discovery suggests the formation of carbonic acid there. This is what I was talking about: it all started with fundamental science and ended with something important for the study of satellites and planets. " .

Note that such pressures turn out to be low relative to those that, in principle, can be found in the Universe, but high in comparison with those that act on us at the Earth's surface.

So sometimes you study something for fundamental science but then you discover it has a right application. For example we have just submitted a paper in which we took carbon, hydrogen, oxygen at high pressure and we tried to look for the all stable compounds. We found one which was carbonic acid and it was stable in a very low pressure like one gigapascal. I investigated the astrophysics literature and discovered: there are satellites such as Ganymede or Calisto. On them there is carbon dioxide and water. The molecules which form this carbonic acid. So we realized that this discovery means that probably there would be carbonic acid. This is what I mean by started for fundamental and discovering something which is applicable to planetary science.

Another example of unusual chemistry that can be cited concerns the well-known table salt, NaCl. It turns out that if you can create a pressure of 350 GPa in your salt shaker, you will get new connections. In 2013, under the leadership of A.R. Oganov showed that if you apply high pressure to NaCl, unusual compounds will become stable - for example, NaCl 7 (Fig. 10) and Na 3 Cl. Interestingly, many of the substances discovered are metals. Gabriele Salekh and Artem Oganov continued their pioneering work in which they showed the exotic behavior of sodium chlorides under high pressure and developed a theoretical model that can be used to predict the properties of alkali metal compounds with halogens.

They described the rules that these substances obey in such unusual conditions. Using the USPEX algorithm, several compounds with the formula A 3 Y (A = Li, Na, K; Y = F, Cl, Br) were theoretically subjected to pressures up to 350 GPa. This led to the discovery of chloride ions in the oxidized state –2. "Standard" chemistry prohibits this. Under such conditions, new substances can form, for example, with the chemical formula Na 4 Cl 3.

Figure 10. Crystal structure of common salt NaCl ( left) and the unusual compound NaCl 7 ( on right) stable under pressure.

Chemistry needs new rules

Gabriele Saleh (Fig. 11) spoke about his research aimed at describing new rules of chemistry that would have predictive power not only under standard conditions, but would describe the behavior and properties of substances under high pressure (Fig. 12).

Figure 11. Gabriele Saleh

“Two or three years ago, Professor Oganov discovered that a salt as simple as NaCl under high pressure is not so simple: sodium and chlorine can form other compounds as well. But nobody knew why. Scientists performed calculations, received results, but it remained unknown why everything happens this way and not otherwise. Ever since graduate school, I have been studying chemical bonding, and in the course of my research I was able to formulate some rules that logically explain what is happening. I studied how electrons behave in the composition of such compounds, and came to the general laws characteristic of them under high pressure. In order to check whether these rules are a figment of my imagination or are still objectively correct, I predicted the structures of similar compounds - LiBr or NaBr and several more similar ones. Indeed, the general rules are followed. In short, I saw that there is the following trend: when you apply pressure to such compounds, they form a two-dimensional metal structure, and then a one-dimensional one. Then, under very high pressure, wilder things start to happen, because the chlorine would then have an oxidation state of -2. All chemists know that chlorine has an oxidation state of -1, this is a typical example from a textbook: sodium loses an electron, and chlorine takes it away. Therefore, the oxidation numbers are +1 and −1, respectively. But that's not how it works under high pressure. We have shown this using some approaches to the analysis of chemical bonds. Also, in the course of my work, I looked for special literature in order to understand if anyone had already observed such patterns. And it turned out that yes, we did. If I am not mistaken, sodium bismuthate and some other compounds obey the described rules. Of course, this is just the beginning. When the following papers on the topic are published, we will find out if our model has real predictive power. Because this is exactly what we are looking for. We want to describe the chemical laws that would be observed even at high pressures " .

Two or three years ago professor Oganov discovered that the simple salt NaCl at high pressure is not very simple and other compounds will form. But nobody know why. They made a calculation they got the results but you cannot say why this is happening. So since during my PhD I specializing in the study of chemical bonding, I investigated this compounds and I find some rule to rationalize what is going on. I investigated how electrons behave in this compounds and I came up with some rules which this kinds of compounds will follow at high pressure. To check whether my rules were just my imagination or they were true I predicted new structures of similar compounds. For example LiBr or NaBr and some combinations like this. And yes, these rules turn out to be followed. In short, just not to be very specialistic, I’ve seen that there is a tendency: when you compress them they would form two-dimensional metals, then one-dimensional structure of metal. And then at very high pressure some more wild would happen because the Cl in this case will have the oxidation number of −2. All the chemist know that the lowest oxidation number of Cl is −1, which is typical textbook example: sodium loses electron and chlorine gets it. So we have +1 and −1 oxidation numbers. But at a very high pressure it is not true anymore. We demonstrated this with some approaches for chemical bonding analysis. In that work also I tried to look at the literature to see if somebody have seen this kind of rules before. And yes, it turned out that there were some. If I’m not mistaken, Na-Bi and other compounds turned out to follow these rules. It is just a starting point, of course. The other papers will come up and we will see whether this model has a real predictive power. Because this is what we are looking for. We want to sketch the chemistry which will work also for high pressure.

Figure 12. The structure of a substance with the chemical formula Na 4 Cl 3, which is formed at a pressure of 125-170 GPa, which clearly demonstrates the emergence of "strange" chemistry under pressure.

If you experiment, then selectively

Despite the fact that the USPEX algorithm is distinguished by great predictive power within the framework of its tasks, the theory always requires experimental verification. The laboratory of computer-aided design of materials is theoretical, as even its name suggests. Therefore, experiments are carried out in collaboration with other research teams. Gabriele Saleh comments on the research strategy adopted in the laboratory as follows:

“We do not conduct experiments - we are theorists. But we often collaborate with people who do it. In fact, I think it's generally difficult. Today science is highly specialized, so it's not easy to find someone who deals with both " .

We don’t do experiments, but often we collaborate with some people who do experiments. Actually I think in fact it's hard. Nowadays the science is very specialized so it's hard to find somebody who does both.

One of the clearest examples is the prediction of transparent sodium. In 2009 in the magazine Nature the results of the work carried out under the leadership of Artem Oganov were published. In the article, the scientists described a new form of Na, in which it is a transparent non-metal, becoming a dielectric under pressure. Why it happens? This is due to the behavior of valence electrons: under pressure, they are displaced into the voids of the crystal lattice formed by sodium atoms (Fig. 13). In this case, the metallic properties of the substance disappear and the qualities of a dielectric appear. A pressure of 2 million atmospheres makes sodium red, and 3 million atmospheres makes it colorless.

Figure 13. Sodium under pressure over 3 million atmospheres. In blue the crystal structure of sodium atoms is shown, orange- bunches of valence electrons in the voids of the structure.

Few believed that classical metal could exhibit this behavior. However, in collaboration with physicist Mikhail Eremts, experimental data were obtained that fully confirmed the prediction (Fig. 14).

Figure 14. Photographs of the Na sample obtained with a combination of transmitted and reflected illumination. Different pressures were applied to the sample: 199 GPa (transparent phase), 156 GPa, 124 GPa, and 120 GPa.

We need to work with a twinkle!

Artem Oganov told us what requirements he makes to his employees:

“First, they must have a good education. Second, be hard-working. If a person is lazy, then I will not hire him, and if I suddenly hire him by mistake, he will be kicked out. Several employees who turned out to be lazy, inert, amorphous, I simply fired. And I think that this is absolutely correct and good even for the person himself. Because if a person is not in his place, he will not be happy. He needs to go to the place where he will work with fire, with enthusiasm, with pleasure. And this is good for the laboratory and good for humans. And those guys who really work beautifully, with a twinkle, so we pay a good salary, they go to conferences, they write articles that are then published in the best world magazines, they will be fine. Because they are in place and because the laboratory has good resources to support them. That is, the guys do not need to think about extra work in order to survive. They can concentrate on science, on their favorite business, and do it successfully. We have some new grants now, and this opens up the opportunity for us to hire a few more people. There is a competition all the time. People apply all year round, of course, I do not take all of them. "... (2016). Crystalline hydrate of 4-aminopyridine, a method for its production, a pharmaceutical composition and a method of treatment and / or prophylaxis based on it. Phys. Chem. Chem. Phys. 18 , 2840–2849;

- Let's deal with the computer design of new materials. First, what is it? Area of ​​expertise? When does the idea and this approach come up?

- The region is quite new, it is only a few years old. In itself, the computer design of new materials has been the dream of researchers, technologists, fundamental scientists for many decades. Because the process of discovering a new material with the properties you need usually takes many years or even decades of work for entire institutes and laboratories. This is a very costly process and you can get frustrated at the end. That is, you are not always able to invent such material. But even when you are successful, success can take many years of work. This does not suit us at all now, we want to invent new materials, new technologies as quickly as possible.

- Can you give an example of such material that does not work out or did not work out to invent?

- Oh sure. For example, for many decades people have been trying to come up with a material harder than diamond. There have been hundreds of publications on this topic. In some of them, people claimed that the material was found harder than diamond, but then inevitably, after some time (usually not very long), these statements were refuted, and it turned out that it was an illusion. Until now, no such material has been found, and it is perfectly clear why. With the help of our methods, we were able to show that this is fundamentally impossible, so there is nothing even to waste time.

- And if you try to simply explain why not?

- A property such as hardness has a finite limit for each given material. If we take all the materials that we can take, then it turns out that there is a kind of global upper limit. It just so happens that this upper limit corresponds to a diamond. Why exactly a diamond? Because in this structure, several conditions are simultaneously fulfilled: very strong chemical bonds, a very high density of these chemical bonds, and they are evenly distributed in space. There is no one direction that is much harder than another, it is a very solid substance in all directions. The same graphite, for example, has stronger bonds than diamond, but all these bonds are located in the same plane, and very weak bonds interact between the planes, and this weak direction makes the entire crystal soft.

- How did the method develop and how did scientists try to improve it?

- The great Edison said, I think, in connection with his invention of the incandescent light bulb: "I have not failed ten thousand times, but only found ten thousand ways that do not work." This is the traditional style of searching for new materials, which is called Edison's in the scientific literature. And from this method, of course, people have always wanted to move away, because it requires a rare Edison luck and Edison patience. And a lot of time as well as money. This method is not very scientific, it is, rather, a scientific "tyk". And people always wanted to move away from this. When computers appeared and they began to solve more or less complex problems, the question immediately arose: "Is it possible to sort all these combinations of different conditions, temperature, pressure, chemical potentials, chemical composition on a computer instead of doing it in a laboratory?" At first, hopes were very high. People looked at this a little optimistic and euphoric, but soon all these dreams crashed into everyday life. In principle, nothing can be achieved by the methods by which people tried to solve the problem.

- Why?

- Because there are infinitely many options for different arrangement of atoms in the structure of a crystal, and each of them will have completely different properties. For example, diamond and graphite are one and the same substance, and due to the fact that the structure is different, their properties are radically different. So there can be infinitely many different options that differ from both diamond and graphite. Where do you start? Where will you stop? How long will this last? And if you also introduce a variable of chemical composition, then you can also think of an infinite number of different chemical compositions, and the task becomes unbearably difficult. Very quickly people realized that traditional, standard methods of solving this problem lead absolutely nowhere. This pessimism completely buried the first hopes that people cherished since the 60s.

- Computer design is still thought of, or at least felt as a visual thing. As I understand it, in the 60s, 70s or 80s this is still not a visual solution, but a mathematical one, that is, it is a faster calculation, calculation.

- As you understand, when you get numbers on the computer, you can always visualize them, but this is not the only point.

- In general, this is a question only of the readiness of the equipment to do this.

- Yes. Numerical counting is primary, because from numbers you can always make a picture, and from a picture, probably numbers too, although not very accurate. There were a number of famous publications from the mid-80s to the mid-90s that finally instilled pessimism in our area. For example, there was a wonderful publication in which it was said that even such simple substances as graphite or ice are absolutely impossible to predict. Or there was an article called "Are Crystals Predictable" and the first word in that article was "no."

- What do you mean "are predictable"?

- The task of predicting crystal structure is at the core of the entire field of new materials design. Since the structure determines the properties of a substance, then in order to predict a substance with the desired properties, it is necessary to predict the composition and structure. The task of predicting the crystal structure can be formulated as follows: suppose that we have specified the chemical composition, suppose it is fixed, for example, carbon. What is the most stable form of carbon under given conditions? Under normal conditions, we know the answer - it will be graphite; at high pressures, we also know the answer - it's a diamond. But creating an algorithm that can give you this turns out to be a very difficult task. Or you can formulate the task in a different way. For example, for the same carbon: what will be the hardest structure corresponding to this chemical composition? It turns out a diamond. And now let's ask another question: what will be the densest? It seems that it is also a diamond, but not. It turns out that a form of carbon denser than diamond can be invented, at least on a computer, and in principle it can be synthesized. Moreover, there are many such hypothetical forms.

- Even so?

- Even so. But nothing harder than diamond comes out. People have learned to get answers to such questions quite recently. More recently, algorithms have appeared, programs have appeared that can do this. In this case, in fact, this entire area of ​​research turned out to be related to our work in 2006. After that, many other researchers also began to tackle this problem. In general, we still do not miss the palm and come up with more and more new methods, new and new materials.

- "Who are "we?

- This is me and my students, graduate students and researchers.

- To make it clear, because “we” is so polysemantic, in this case polysemantic, it can be perceived in different ways. And what's so revolutionary?

- The fact is that people realized that this task is associated with an infinitely complex combinatorial problem, that is, the number of options among which you need to choose the best is infinite. How can this problem be solved? No way. You can simply not approach her and feel comfortable. But we have found a way that this problem can be solved efficiently enough - a way based on evolution. This, one might say, is a method of successive approximations, when from initially weak solutions by the method of successive improvement we come to more and more perfect solutions. We can say that this is an artificial intelligence method. Artificial intelligence, which makes a number of assumptions, rejects some of them, and constructs even more interesting ones from the most plausible, most interesting structures and compositions. That is, he learns from his own history, which is why it can be called artificial intelligence.

- I would like to understand how you invent, invent new materials using a specific example.

- Let's try to describe it using the example of the same carbon. You want to predict which form of carbon is the hardest. A small number of random carbon structures are specified. Some structures will consist of discrete molecules, like fullerenes; some structures will consist of layers, like graphite; some will consist of carbon chains, the so-called carbines; some will be three-dimensionally connected, like diamond (but not only diamond, there are infinitely many such structures). You first generate these kinds of structures randomly, then you do local optimization, or what we call "relaxation". That is, you move the atoms until the resultant force on the atom is zero, until all stresses in the structure disappear, until it enters its ideal form or gains its best local shape. And for this structure, you calculate properties such as hardness. We look at the hardness of fullerenes. There are strong bonds, but only within the molecule. The molecules themselves are very weakly bound to each other, due to which the hardness is practically zero. Look at graphite - the same story: strong bonds within a layer, weak between layers, and as a result, the substance disintegrates very easily, its hardness will be very low. Substances such as fullerenes or carbines or graphite will turn out to be very soft and we discard them immediately. The remaining carbon structures are three-dimensionally connected, they have strong bonds in all three dimensions, from these structures we choose the most solid ones and give them the opportunity to produce daughter structures. What does it look like? We take one structure, we take another structure, cut out their pieces, put them together, as in a constructor, and again we relax, that is, we let all the stresses go away. There are mutations - this is another way of producing offspring from parents. We take one of the hardest structures and mutate it, for example, we apply a huge shear stress so that some connections there simply burst, and others, new ones, are formed. Or we shift the atoms in the weakest directions of the structure in order to remove this weakness from the system. We relax all the structures produced in this way, that is, we remove internal stresses, and then we re-evaluate the properties. It so happens that we took a solid structure, mutated it, and it became soft, turned into, say, graphite. We immediately remove such a structure. And from those that are solid, we again produce "children". And so we repeat it step by step, generation after generation. And quickly enough we come to the diamond.

- At the same time, the moments when we reject, compare, combine and change the structure, does artificial intelligence do it, does the program? Not a human?

- The program does it. If we did this, we would end up in Kashchenko, because this is a huge number of operations that a person does not need to do and for completely scientific reasons. You understand, a person is born, absorbs experience from the world around him, and with this experience comes a kind of prejudice. We see a symmetrical structure - we say, "This is good"; we see asymmetrical - we say: "This is bad." But for nature, it sometimes happens the other way around. Our method must be free from human subjectivity and prejudice.

- Do I understand correctly from what you described that, in principle, this task is formulated not so much by fundamental science as by the solution of very specific tasks set by some regular transnational company? So we need a new cement so that it is more viscous, denser, or, conversely, more liquid, and so on.

- Not at all. In fact, I came from fundamental science by my education, I still studied fundamental science, not applied science. I am now interested in solving applied problems, especially since the methodology that I have invented is applicable to the most important applied problems of a very wide range. But initially this method was invented to solve fundamental problems.

- What kind of?

- I have been studying physics and chemistry of high pressures for a long time. This is an area in which many interesting discoveries have been made experimentally. But experiments are difficult, and very often the experimental results have turned out to be wrong over time. Experiments are costly and time consuming.

- Give an example.

- For example, for a long time there was a race between Soviet and American scientists: who will get the first metallic hydrogen under pressure. Then it turned out, for example, that many simple elements under pressure become (this is such an alchemical transformation) a transition metal. For example, you take potassium: potassium has only one s-electron on its valence shell, and under pressure it becomes a d-element; The s orbital is emptied, and the unoccupied d orbital is populated by this single electron. And this is very important, because potassium, becoming a transition metal, then gets the opportunity to enter, for example, liquid iron. Why is it important? Because now we believe that potassium in small quantities is included in the core of the Earth and is a source of heat there. The fact is that one of the isotopes of potassium (radioactive potassium-40) is one of the main producers of heat on Earth today. If potassium is not included in the Earth's core, then we must completely change our understanding of the age of life on Earth, the age of the magnetic field, the history of the Earth's core and many other interesting things. Here is an alchemical transformation - s-elements become d-elements. At high pressures, when you compress a substance, the energy that you spend on compression will sooner or later exceed the energy of a chemical bond and the energy of interorbital transitions in atoms. And thanks to this, you can radically change the electronic structure of the atom and the type of chemical bond in your substance. Completely new types of substances can arise. And the standard chemical intuition in such cases does not work, that is, the rules that we learn from school in chemistry lessons, they fly to hell when the pressure reaches sufficiently high values. I can tell you what kind of things were predicted by our method and then experimentally proven. When this method appeared, it came as a shock to everyone. One of the most interesting works involved the element sodium. We predicted that if sodium is compressed to a pressure of about 2 million atmospheres (by the way, the pressure in the center of the Earth is almost 4 million atmospheres, and such pressures can be obtained experimentally), it will no longer be a metal, but a dielectric, moreover, transparent and red colors. When we made this prediction, no one believed us. The journal Nature, to which we sent these results, even refused to consider this article, they said that it was impossible to believe it. I contacted the experimenters from the group of Mikhail Eremts, who also told me that it was impossible to believe in this, but out of respect they would still try to conduct such an experiment. And this experiment fully confirmed our predictions. The structure of the new phase of the element boron was predicted - the hardest structure for this element, one of the hardest substances known to mankind. And there it turned out that different boron atoms have different electric charges, that is, they suddenly become different: some positively, some negatively charged. This article has been cited nearly 200 times in just three years.

- You said that this is a fundamental task. Or do you primarily solve fundamental problems and only recently - some practical issues? The sodium story. What for? That is, you sat, sat and thought, what to take - I will take sodium, perhaps, and squeeze it into 2 million atmospheres?

- Not certainly in that way. I received a grant to study the behavior of elements under high pressure in order to better understand the chemistry of elements. Experimental data under high pressure is still very fragmentary, and we decided to go through more or less the entire Periodic Table to understand how the elements and their chemistry change under pressure. We have published a number of articles, in particular, on the nature of superconductivity in oxygen under pressure, because oxygen under pressure becomes a superconductor. For a number of other elements: alkaline elements or alkaline earth elements, and so on. But the most interesting, probably, was the discovery of new phenomena in sodium and boron. These were probably the two elements that surprised us the most. This is how we started. And now we have moved on to solving practical problems, we cooperate with companies such as Intel, Samsung, Fujitsu, Toyota, Sony. Toyota, as far as I know, has recently invented a new material for lithium batteries using our method and is going to bring this material to the market.

- They took your method, took the technology of finding materials, but not you?

- Oh sure. We do not impose ourselves on the load, but try to help all researchers. Our program is available to everyone who wants to use it. Companies need to pay something for the right to use the program. And scientists working in academic science get it for free by simply downloading it from our website. Our program already has almost 2 thousand users all over the world. And I am very happy when I see that our users are achieving something good. I, my group have more than enough of my discoveries, my works, my insights. When we see the same thing in other groups, it only makes us happy.

The material was prepared on the basis of the radio broadcast "PostNauka" on the radio Russian News Service.

1. 1. Computer design of new materials: dream or reality? Artem Oganov (ARO) (1) Department of Geosciences (2) Department of Physics and Astronomy (3) New York Center for Computational Sciences State University of New York, Stony Brook, NY 11794-2100 (4) Moscow State University, Moscow, 119992, Russia.
2. 2. The structure of matter: atoms, molecules The ancients guessed that matter consists of particles: “when He (God) did not create either the earth, or fields, or the initial dust particles of the universe” (Proverbs 8:26) (also - Epicurus, Lucretius Car , ancient Indians, ...) In 1611, I. Kepler suggested that the structure of the ice-form of snowflakes is determined by their atomic structure
3. 3. The structure of matter: atoms, molecules, crystals 1669 - the birth of crystallography: Nikolai Stenon formulates the first quantitative law of crystallography “Crystallography ... is unproductive, exists only for itself, has no consequences ... not being really needed anywhere, it developed inside yourself. It gives the mind some limited satisfaction, and its details are so varied that it can be called inexhaustible; that is why it lassoes even the best people so tenaciously and for so long "(IV Goethe, amateur crystallographer, 1749-1832) Ludwig Boltzmann (1844-1906) - the great Austrian physicist who built all his theories on the concepts of atoms. Criticism of atomism led him to commit suicide in 1906. In 1912, the hypothesis of the atomic structure of matter was proved by the experiments of Max von Laue.
4. 4. Structure - the basis for understanding the properties and behavior of materials (from http://nobelprize.org) Zinc blende ZnS. One of the first structures solved by Braggs. In 1913 Surprise: there are NO ZnS molecules in the structure!
5. 5. X-ray diffraction is the main method for the experimental determination of the crystal structure. Structure Diffraction pattern
6. 6. Relationship between structure and diffraction pattern What will be the diffraction patterns of these "structures"?
7. 7. Triumphs of experiment - determination of incredibly complex crystal structures Inappropriate phases Quasicrystals of elements Proteins (Rb-IV, U.Schwarz'99) A new state of matter, discovered in 1982. Found in nature only in 2009! 2011 Nobel Prize!
8. 8. States of matter Crystalline Quasicrystalline Amorphous Liquid Gaseous ("Soft matter" - polymers, liquid crystals)
9. 9. Atomic structure is the most important characteristic of a substance. Knowing it, one can predict the properties of the material and its electronic structure. Theory Exp. C11 493 482 C22 546 537 C33 470 485 C12 142 144 C13 146 147 C23 160 146 C44 212 204 C55 186 186 Elastic constants of MgSiO3 perovskite C66 149 147
10. 10. Several stories 4. Materials of the earth's interior 3. Materials from the computer 2. Is it possible to predict crystalline1. On the relationship between structure, structure and properties
11. 11. Why ice is lighter than water The structure of ice contains large empty channels that are not present in liquid water. Because of these empty channels, ice is lighter than ice.
12. 12. Gas hydrates (clathrates) - ice filled with guest molecules (methane, carbon dioxide, chlorine, xenon, etc.) Number of publications on clathrates Huge deposits of methane hydrate - the hope and salvation of the energy sector? Under low pressure, methane and carbon dioxide form clathrates - 1 liter of clathrate contains 168 liters of gas! Methane hydrate looks like ice, but burns with the release of water. CO2 hydrate - a form of carbon dioxide storage? The mechanism of xenon anesthesia is the formation of He-hydrate, which blocks the transmission of neural signals to the brain (Pauling, 1951)
13. 13. Microporous materials for the chemical industry and environmental clean-up Zeolites are microporous aluminosilicates. industry Historical examples of heavy metal poisoning: Qin Shi Huangdi Ivan IV the Terrible "Disease Nero (37-68) Lead (259 - 210 BC) (1530-1584) mad poisoning: hatter" aggression, dementia
14. 14. New and old superconductors The phenomenon was discovered in 1911 by Kamerling-Onnes. The theory of superconductivity is 1957 (Bardeen, Cooper, Schrieffer), but there is no theory of the highest-temperature superconductors (Bednorz, Muller, 1986)! The most powerful magnets (MRI, mass spectrometers, particle accelerators) Trains with magnetic levitation (430 km / h)
15. 15. Surprise: superconducting impurity forms of carbon 1.14 1 Tc  exp [] kB g (E F) V Doped graphite: KC8 (Tc = 0.125 K), CaC6 (Tc = 11 K). B-doped diamond: Tc = 4 K. Doped fullerenes: RbCs2C60 (Tc = 33 K) Molecule molecule Structure and appearance of fullerene C60 fullerite crystals Superconductivity in organic crystals has been known since 1979 (Bechgaard, 1979).
16. 16. How materials can save or destroy At low temperatures, tin undergoes a phase transition - "tin plague". 1812 - according to legend, Napoleon's expedition to Russia died because of the tin buttons on their uniforms! 1912 - the death of the expedition of Captain R.F. Scott to the South Pole, which was attributed to the "tin plague." First order transition at 13 0C White tin: 7.37 g / cm3 Gray tin: 5.77 g / cm3
17. 17. Shape memory alloys 1 2 3 4 1- before deformation 3- after heating (20 ° C) (50 ° C) 2- after deformation 4- after cooling (20 ° C) (20 ° C) Example: NiTi ( nitinol) Applications: Shunts, dental braces, elements of oil pipelines and aircraft engines
18. 18. Wonders of optical properties Pleochroism (cordierite) - the discovery of America and navigation of the US Air Force Birefringence of light (calcite) Alexandrite effect (chrysoberyl) Lycurgus bowl (glass with nanoparticles)
19. 19. About the nature of color Wavelength, Å Color Additional color 4100 Violet Lemon yellow 4300 Indigo Yellow 4800 Blue Orange 5000 Blue-green Red 5300 Green Purple 5600 Lemon yellow Violet 5800 Yellow Indigo 6100 Orange Blue 6800 Red Blue-green
20. 20. Color depends on direction (pleochroism). Example: cordierite (Mg, Fe) 2Al4Si5O18.
21. 21. 2. Prediction of crystal structures Oganov A.R., Lyakhov A.O., Valle M. (2011). How evolutionary crystal structure prediction works - and why. Acc. Chem. Res. 44, 227-237.
22. 22. J. Maddox (Nature, 1988) The task is to find the GLOBAL minimum of Natoms of Variants Time of energy. 1 1 1 sec. It is impossible to iterate over all structures: 10 1011 103 yrs. 20 1025 1017 yrs. 30 1039 1031 yrs. Review of the USPEX method (ARO & Glass, J. Chem. Phys. 2006)
23. 23. How to find Mount Everest with the help of kangaroo evolution? (picture from R. Clegg) We are landing kangaroos and letting them breed (not shown for censorship reasons) .....
24. 24. How to find Mount Everest with the help of kangaroo evolution? (picture from R. Clegg) Aaaargh! Ouch .... and from time to time hunters come and remove kangaroos at lower altitudes
25. 25.
26. 26. Evolutionary calculations "self-learn" and focus the search on the most interesting areas of space
27. 27. Evolutionary calculations "self-learn" and focus the search on the most interesting areas of space
28. 28. Evolutionary calculations "self-learn" and focus the search on the most interesting areas of space
29. 29. Evolutionary calculations "self-learn" and focus the search on the most interesting areas of space
30. 30. Alternative Methods: Random Search (Freeman & Catlow, 1992; van Eijck & Kroon, 2000; Pickard & Needs, 2006) No "learning", only works for simple systems (up to 10-12 atoms). Artificial annealing (Pannetier 1990 ; Schön & Jansen 1996) No “learning” Metadynamics (Martonak, Laio, Parrinello 2003) Taboo search in reduced space Minima hopping (Gödecker 2004) Uses calculation history and “self-learning”. Genetic and evolutionary algorithms Bush (1995), Woodley (1999) - ineffective method for crystals. Deaven & Ho (1995) is an efficient method for nanoparticles.
31. 31. USPEX (Universal Structure Predictor: Evolutionary Xtallography) (Random) initial population New generation of structures is produced only from the best current structures (1) Heredity (3) Coordinate (2) Lattice mutation mutation (4) Permutation
32. 32. Additional techniques - the parameter of order "Fingerprint" of the structure The birth of order from chaos in the evolutionary process ["GOD = Generator Of Diversity" © S. Avetisyan] Local order - indicates defective areas
33. 33. Test: “Who would guess that graphite is the stable allotrope of carbon at ordinary pressure?” (Maddox, 1988) Three-dimensional sp2 structure proposed by R. Hoffmann (1983) as a stable phase at 1 atm. Structures with low sp3- energy hybridization illustrate sp2 hybridization carbon chemistry sp hybridization (carbyne)
34. Test: High pressure phases are also reproduced correctly 100 GPa: diamond is stable 2000 GPa: bc8 phase is stable + metastable phase found to explain the metastable bc8 phase of silicon “superhard graphite” is known (Kasper, 1964) (Li, ARO, Ma, et al., PRL 2009)
35. 35. Discoveries made with USPEX:
36. 36. 3. Materials from the computer
37. 37. Discovery of new materials: still an experimental method of trial and error "I did not suffer (ten thousand) failures, but only discovered 10,000 non-working methods" (T.A. Edison)
38. 38. Search for the densest substance: are carbon modifications denser than diamond possible? Yes Diamond structure Diamond has the smallest atomic volume and the highest incompressibility among all new structure, elements (and compounds). denser than diamond! (Zhu, ARO, et al., 2011)
39. 39. The analogy between the forms of carbon and silica (SiO2) allows us to understand the density of new forms of carbon New structures, 1.1-3.2% denser than diamond, very high (up to 2.8!) Refractive indices and light dispersion diamond hP3 structure tP12 structure tI12 structure SiO2 cristobalite SiO2 quartz SiO2 kitite high pressure SiS2 phase
40. 40.
41. 41. The hardest oxide - TiO2? (Dubrovinsky et al., Nature 410, 653-654 (2001)) Nishio-Hamane (2010) and Al-Khatatbeh (2009): compression modulus ~ 300 GPa, not 431 GPa. Lyakhov & ARO (2011): Pressure experiments are very difficult! Hardness not higher than 16 GPa! TiO2 is softer than SiO2 of stishovite (33 GPa), B6O (45 GPa), and Al2O3 of corundum (21 GPa).
42. 42. Are forms of carbon harder than diamond possible? No . Material Model Li Lyakhov Exp. Hardness, Enthalpy, et al. & ARO Structure GPa eV / atom (2009) (2011) Diamond 89.7 0.000 Diamond 91.2 89.7 90 Lonsdaleite 89.1 0.026 Graphite 57.4 0.17 0.14 C2 / m 84.3 0.163 TiO2 rutile 12.4 12.3 8-10 I4 / mmm 84.0 0.198 β-Si3N4 23.4 23.4 21 Cmcm 83.5 0.282SiO2 stishovite 31.8 30.8 33 P2 / m 83.4 0.166 I212121 82.9 0.784 Fmmm 82.2 0.322 Cmcm 82.0 0.224 P6522 81.3 0.111 All the hardest structures are based on sp3 hybridization Evolutionary calculation
43. 43. Cold compression of graphite gives M-carbon, not diamond! M-carbon was proposed in 2006. In 2010-2012. dozens of alternative structures have been proposed (W-, R-, S-, Q-, X-, Y-, Z-carbon, etc.) M-carbon is confirmed by the latest experiments M-carbon is most easily formed from graphite graphite bct4-carbon graphite M -carbon graphite diamond
44. 44. M-carbon - a new form of carbon diamond-graphite lonsdaleite Theoretical phase diagram of carbon M-carbon fullerenes carbines
45. 45. Substance under pressure in nature P.W. Bridgman 1946 Nobel laureate (Physics) 200x Scale: 100 GPa = 1 Mbar =
46. Neptune has an internal heat source - but where does CH4 come from? Uranus and Neptune: H2O: CH4: NH3 = 59: 33: 8. Neptune has an internal energy source (Hubbard'99). Ross'81 (and Benedetti'99): CH4 = C (diamond) + 2H2. Falling diamond - the main source of heat on Neptune? Theory (Ancilotto'97; Gao '2010) confirms this. methane hydrocarbons diamond
47. 47. Boron is located between metals and non-metals and its unique structures are sensitive to B impurities, temperature and pressure alpha-B beta-B T-192
48. 48. The history of the discovery and research of boron is full of contradictions and detective twists and turns. B 1808: J.L. Gay-Lussac and H. Davy announced the discovery of a new element - boron. J.L. Gay-Lussac H. Davy 1895: H. Moissan proved that the substances they discovered contain no more than 50-60% boron. The Moissan material, however, also turned out to be a compound with a boron content of less than 90%. H. Moissan 1858: F. Wöhler described 3 modifications of boron - "diamond", "graphite" and "carbon-like". All three turned out to be compounds (for example, AlB12 and B48C2Al). 2007: ~ 16 crystal modifications were published (are most of them compounds?). It is not known which form is the most stable. F. Wöhler
49. 49. Boron forms a partially ionic structure under pressure! B 2004: Chen and Solozhenko: synthesized a new boron modification, but could not solve its structure. 2006: Oganov: defined the structure, proved its stability. 2008: Solozhenko, Kurakevich, Oganov - this phase is one of the hardest known substances (hardness 50 GPa). X-ray diffraction. Top - theory, Bottom - experiment Structure of gamma boron: (B2) δ + (B12) δ-, δ = + 0.5 (ARO et al., Nature 2009) Distribution of the most (left) and least (right) stable electrons.
50. 50. The first boron phase diagram - after 200 years of research! B Phase diagram of boron (ARO et al., Nature 2009)
51. 51. Sodium is a metal perfectly described by the free electron model
52. 52. Under pressure sodium changes its essence - "alchemical transformation" Na 1807: Sodium is discovered by Humphrey Davy. 2002: Hanfland, Syassen, et al. - the first indication of extremely difficult chemistry H. Davy sodium under pressure over 1 Mbar. Gregoriants (2008) - more detailed data. Under pressure, sodium becomes partially d-metal!
53. 53. We have predicted a new structure that is a transparent non-metal! Sodium becomes transparent at a pressure of ~ 2 Mbar (Ma, Eremets, ARO et al., Nature 2009) Electrons are localized in the “empty space” of the structure, which makes compressed sodium a non-metal
54. The study of minerals is not only aesthetic pleasure, but also a practically and fundamentally important scientific direction Effect of lowering the melting point of impurities Wood's alloy - melts at 70 C. Bi-Pb-Sn-Cd-In-Tl alloy - at 41.5 C!
55. 64. And what is the composition of the inner core of the Earth? The core is somewhat less dense than pure iron. In the Fe core in an alloy with light elements such as S, Si, O, C, H. New compounds (FeH4!) Are predicted in the Fe-C and Fe-H systems. Carbon can be contained in the core in large quantities [Bazhanova, Oganov, Dzhanola, UFN 2012]. The percentage of carbon in the inner core needed to explain its density
56. 65. The nature of the D "layer (2700-2890 km) has long remained a mystery. D" is the root of hot mantle flows. MgSiO3 is expected to be ~ 75 vol.% Strangeness of the D "layer: seismic rupture, anisotropy Recall the anisotropy of cordierite color!
57. 66. The answer is the existence of a new mineral, MgSiO3 post-perovskite in layer D "(2700-2890 km) Phase diagram D" rupture of MgSiO3 Explains the existence of layer D ", allows to calculate its temperature. perovskite as the Earth cools D “is ​​absent on Mercury and Mars New family of minerals predicted Confirmation - Tschauner (2008)
58. 67. The structure of matter is the key to understanding the world 4. The understanding of the planetary interior deepens 3. The computer learns to predict new materials 2. It is already possible to predict crystal structures1. Structure defines properties
59. 68. Acknowledgments: My students, graduate students and postdocs: A. Lyakhov Y. Ma S.E. Boulfelfel C.W. Glass Q. Zhu Y. Xie Colleagues from other laboratories: F. Zhang (Perth, Australia) C. Gatti (U. Milano, Italy) G. Gao (Jilin University, China) A. Bergara (U. Basque Country, Spain) I. Errea (U. Basque Country, Spain) M. Martinez-Canales (UCL, UK) C. Hu (Guilin, China) M. Salvado & P. ​​Pertierra (Oviedo, Spain) V.L. Solozhenko (Paris) D.Yu. Pushcharovsky, V.V. Brazhkin (Moscow) USPEX program users (> 1000 people) - http://han.ess.sunysb.edu/~USPEX