| |
It appears as though we finally have the tools available to begin to quickly and accurately model the atomic structure of bulk materials and to predict their associated properties. This represents a significant advance and one that I'm surprised took so long to achieve.
A recent article(1) from NewScientist introduces the work of Chris Pickard (University College London), who developed a computer program that models the bonds and forces between different atoms in a material and predicts the properties that the material will have. His latest program has performed very well in predicting crystal structures. Genetic algorithms have also been applied to modelling materials and predicting their properties for a number of years now. As an example, Artem Organov (Stoney Brook University, New York), used improved genetic algorithms to predict that, under 3 million atmospheres pressure, pure sodium turns from a red metal to a transparent insulator; subsequent experiments confirmed that this is indeed what happens. I've always been a big fan of genetic or evolutionary algorithms and their ability to rapidly find creative solutions to problems that a person would rarely discover.
The current suite of material modelling programs appears to be restricted to the user specifying the type and general arrangement of atoms and then predicting what the bulk material will have under different conditions in different environments.
The real promise of these programs however, or at least of the gext generation of programs, is to be able to do the reverse. To specify to the program the properties of the material that you would like, and then for the modelling program to explore a large computational space and determine the ideal atomic identity and structure required for a bulk material that would possess such properties.
For example, some properties that we may like to specify would include room-temperature superconductivity, ultra-hardness, super-efficient (90%+) solar to electric energy conversion, optimised metamaterials for optical manipulation and control (e.g. invisibility materials, super lenses, etc), opimised engineerable self-assembly materials (improved DNA origami), extreme-capacity hydrogen storage materials, etc. With enough processing capacity such modelling software would quickly be able to determine the ideal atomic structures that are required by a material in order to possess such properties.
This step is essentially the easy part; the hard part is actually being able to manufacture the resulting materials in economically useful quantities.
But with such concrete and easily visible goals for people to aim at, mobilising the resources required to design suitable manufacturing and processing methods should be easy. Do you really think an Intel or a Toshiba wouldn't throw $50 Million at the R&D required to produce such materials? The pay-off, the ROI on such a project would be astronomical.
The next steps should be:
1. Provide a significant boost to basic research in this area of computational materials modelling, spread over the leading groups around the world. This would enable further development of these algorithms, and also importantly, go towards proving that the predicted properties of materials that these programs make are true across every variable and property.
2. The programs should not only predict, from basic principles, that HgBa2Ca2Cu3O8+δ superconducts at 138K, but also determine the transition temperature and other properties of many other variants (that are manufacturable with todays methods but haven't yet been so) that can then be made and tested.
3. The best performing, most accurate programs can then be used to predict the elemental structure and composition of all the "Super-materials" that our present technological civilisation desires; things like room temperature superconductors, extremely efficient solar cells, etc, as mentioned above.
4. Significant commerical development funding would then come on board to design the manufacturing processes required to produce the materials in bulk, with subsequent commercialisation improving the standard of living for all.
5. Later generations of the predictive modelling programs would optimise other aspects of these materials such as manufacturability, durability, etc.
6. As computational power increases and with the addition of specialised AI capabilities, the breadth and scope of these programs would increase substantially. Instead of just designing the optimal elemental configuration of materials these programs would additionally design these into components and then into fully integrated systems and products and even the new tools required to manufacture these items; hopefully just in time for nano-molecular manufacturing.
The power of such an approach cannot be overstated.
(1) http://www.newscientist.com/article/mg20427301.200 -solving-the-crystal-maze-the-secrets-of-structure .html?page=1 |
