Algorithm Predicts New Superhard Materials

An anonymous reader writes "Researchers in New York have developed an algorithm that can predict new superhard materials — a relatively small class of compounds of which diamond is the most famous. Beyond the pluses this represents for, say, the drilling industry, the physicists claim say their computational approach can be used to think up new materials of all sorts. 'New materials with desired properties will be routinely discovered using supercomputers,' they say, 'instead of the expensive trial-and-error method that is used today.'"

Comments

Re:Penis?

[Farmer Tim]

Yes, nanotubes are quite hard.

Project - Mc Lab / Magic Chemist, in a Box.

[John Sokol]

I wrote up a plan for something like this about 2 1/2 years ago and posted on my blog about 9 months ago when it became obvious to me that as cool of an idea as it was, it wasn't something I wanted to work on.

The basic idea is to take a computational chemistry package and run it through a genetic algorithm to search for suitable candidates that solve certain problems.
Better solar cells, dielectrics for supercaps, or materials with specific properties.

The physics quickly went over my head and I was never able to get funding or grants for this without a PhD.

I am glad to see this is starting to happen.

Project - Mc Lab / Magic Chemist, in a Box.
  http://johnsokol.blogspot.com/2010/12/project-mc-lab-magic-chemist-in-box.html
  http://thegreentank.blogspot.com/2010/12/project-mc-lab-magic-chemist-in-box.html

Re:Project - Mc Lab / Magic Chemist, in a Box.

[John Sokol]

Well with enough input knowledge of molecules. You could also use Neural networks or GA to evolve better models, but I did realize the problem you are referring to.

Again it's not going to be 100%, maybe not even 50% but even 10% would still reduce the search space immensely. The downside is you could easily overlook optimal solutions that don't model correctly.