The inverse problem in Materials Design poses some unique challenges. Specifically, when one considers all the possible molecular structures that are possible, the combinatorial complexity of chemical composition space is overwhelming. Additional challenges in the inverse problem result because the composition variables are both discrete (e.g.. molecule A vs. molecule B) and continuous (e.g. concentration) and because the engineering properties are highly nonlinear functions in the composition hyperspace with numerous local minima.  In order to address these challenges, several groups have developed Genetic Algorithm (GA) based search methods.  Genetic Algorithms are stochastic based search methods that are based on the survival-of-the-fittest strategy.  Specifically, (i) an initial population of candidate materials is generated, (ii) progeny are created from the initial population by various operations like crossover and mutation, (iii) the performance of the new population is determined using the forward model and (iv) the best members of the population are allowed to survive to the next generation, while the rest are discarded.

The GA procedure has been very efficient in determining optimal compositions in very large search spaces that are impossible to search by other techniques.  In addition, we are now developing hybrid inverse methods that use expert knowledge to guide the stochastic search.

  Up Coming Event

Summer School on the Materials Genome Project
3-8 September 2007, Oran, Algeria