Upending chemistry’s atomic law

Scientists have found radical means of creating compounds that could transform the world. Uspex, a computer program that exploits the power of Darwinian evolution, seeks to produce new, stable materials by uncovering the best combinations of atoms, Robert Matthews writes

Margarita salt, SEM

It’s not just people that behave strangely under pressure. Pretty much everything will do weird things if when under the cosh.

Crush a sugar cube under a glass tumbler and it’ll emit flashes of faint blue light.

Put the squeeze on hydrogen, and it will turn from the lightest of gases into a solid metal. Do the same to peanut butter and it turns into diamonds.

Scientists have long known the cause of all these weird effects. Pressure distorts the bonds between atoms, forcing them to take up strange new structural arrangements.

But now they’re having to come to terms with something far more radical. It seems pressure can cause even the fundamental laws of chemistry to buckle.

At school we all had to memorise the formulas for various common compounds: H2O for water, CO2 for carbon dioxide and the like. We were also supposed to be able to explain why the formulas had to be this way; something to do with “electron shells”, whatever they were.

But we all knew we would get in trouble if we just tried to guess the formulas. Common salt is NaCl: one atom of sodium, plus one atom of chlorine. Nothing else works, and there is no way around it: it is the laws of chemistry.

Or at least it was until late last month, when researchers announced the creation of a whole family of the mutant relatives of common salt, with crazy formulas like NaCl3 and Na2Cl.

Standard chemistry forbids such compounds, and yet now they have been created in experiments reported in the leading journal Science.

This is no mere party trick, either. It is confirmation of a radically new means of creating compounds that could transform our world.

It is a curious fact that, despite centuries of effort, chemists still struggle to predict how atoms bind together to form new materials.

Such materials need stable crystal structures, and working how the inter-atomic forces create them has proved incredibly difficult. Put simply, it is like the rubik’s cube from hell: there are zillions of different ways of combining atoms, but the vast majority lead nowhere because they are unstable.

Over the last decade or so, however, a team led by Professor Artem Oganov of the State University of New York, Stony Brook, has developed a recipe for homing in on the handful of best candidates buried amid the dross.

Known as Uspex, it is a computer program that performs its magic by exploiting the power of Darwinian evolution.

In essence, evolution is the search for combinations of properties that will allow organisms to stay alive long enough to reproduce.

There is a huge number of possible combinations, most are useless.

But evolution hones in on the better ones by using random mutation and natural selection.

The mutations introduce variety, while natural selection acts as a filter, weeding out those combinations that do not succeed as well as others.

Computer scientists have long used such “genetic algorithms” to tackle problems that have a myriad of mediocre solutions but only a handful of good ones.

For example, finding the quickest route between just 20 towns involves finding the best from the billion billion different possible pathways. A genetic algorithm can do this in seconds by testing an initial set of guesses, slicing and dicing them, mutating and combining the best ones, and then testing the results again.

Logistics companies use such algorithms to find the optimal routes for their deliveries.

Now Prof Oganov and his colleagues are using the same idea to sift through the myriad combinations of atoms to see which ones may lead to stable materials.

Finding them requires some heavy-duty computing power. The team is currently running its software on the Stampede supercomputer at the University of Texas, Austin, which is capable of up to 10 million billion calculations a second – around a million times faster than your laptop. The resulting compounds don’t just have weird formulas though: their predicted properties are often no less bizarre.

One of the first big surprises emerged in 2009, when the Uspex algorithm predicted that the metal sodium would become transparent like glass when exposed to high pressure.

According to the textbooks, all materials, even hydrogen, exposed to high enough pressures should turn into opaque metals. Yet Uspex suggested sodium would do the exact opposite, turning from a white metal to black as the pressure increases, then to red and finally transparent.

At first, the suspicion was that Uspex had goofed, showing its limitations at last. But the only way to know for sure was to try it. So Prof Oganov and his colleagues persuaded Dr Mikhail Eremets at the Max Planck Institute of Chemistry in Mainz, Germany, to set aside his scepticism and put the prediction to the test.

Cranking up the pressure to 1.5 million atmospheres, Dr Eremets and his colleagues saw the sodium turn black. Pushing onwards to 2 million atmospheres and beyond, the sodium turned red and then finally transparent – just as the algorithm had predicted.

Since then, Prof Oganov and others have made many unexpected discoveries in what looks like the dawn of a new era in chemistry.

And if the history of this most practical of sciences is any guide, we can expect this major advance to have a significant impact on our lives.

Some hints of where they might be are already starting to emerge. The “impossible” combination of three sodium atoms and one of chlorine (Na3Cl) announced last month appears to be like an atomic layer-cake, with ordinary salt alternating with layers of pure sodium.

This structure will pique the interest of technologists, as salt does not conduct electricity very well, while sodium does – and having an atomic-scale sandwich of the two could have some practical applications in the electronics and power industries.

Exploiting such properties will require some way of keeping the compounds stable at everyday pressures. Prof Oganov and his colleagues hope to tackle this using Uspex to find suitable tweaks to the atomic structure.

By coincidence, their paper announcing the creation of “impossible” molecules came in the year-end issue of Science, trumpeting the biggest breakthroughs of 2013.

If the team’s algorithm lives up to its promise, it will rate as one of the breakthroughs of the century.

Robert Matthews is a visiting reader in science at Aston University in Birmingham