Hundreds of thousands of possible metal alloy combinations can be formed from a relatively small number of elements. As a result, finding new metal blends with desirable qualities (such as rust resistance or heat conductivity) can be an arduous task. Now a team of Danish physicists has developed a new approach to this treasure hunt that moves the search from the laboratory bench to the desktop. A report published in the current issue of Physical Review Letters describes a computer algorithm that borrows from evolutionary theory to test which compounds hold the most promise for resisting high temperatures and corrosion.

Gisli Jhannesson and colleagues at the Technical University of Denmark cybersifted through 192,016 potential alloys, each containing up to four different metals (out of 32 possible choices), using a genetic algorithm. The researchers started with a so-called living population of alloys with various compositions. The program then created new populations of alloys in one of two ways: breeding between two parent alloys, in which elements were interchanged to form child alloys; and mutations, which replaced one element in an alloy with a randomly selected substitute. The program then selected the most stable alloys according to density functional theory (DFT), which models interactions between electrons to predict a material's properties. By manipulating the candidate metals, the team could account for practical considerations such as the prices of particular elements.

The researchers ran a number of initial populations through their evolutionary computer program and found among their survivors some of the best superalloys yet known. Their list also contained several alloys that have either not been well studied or considered at all as candidates for superalloys. The authors conclude that "the purely computational approach may therefore be able to narrow down the number of experiments needed for the development of new materials."