Molecular nanotechnology is the precise, three-dimensional control of materials and devices at the atomic scale.An important part of nanotechnology is the design of moleculesfor specific purposes. This paper describes early results usinggenetic software thchniques to automatically design moleculesunder the control of a fitness function. The fitness function mustbe capable of determining which of two arbitrary molecules isbetter for a specific task. The software begins by generating apopulation of random molecules. The individual molecules in apopulation are then evolved towards greater fitness by randomlycombining parts of the better existing molecules to create newmolecules. These new molecules then replace some of the lessfitmolecules in the population. We apply a unique genetic crossoveroperator to molecules represented by graphs, i.e., sets of atomsand the bonds that connect them. We present evidence suggestingthat crossover alone, operating on graphs, can evolve any possiblemolecule given an appropriate fitness function and a populationcontaining both rings and chains. Most prior work evolved stringsor trees that were subsequently processed to generate moleculargraphs. In principle, genetic graph soft ware should be able toevolve other graph-representable systems such as circuits,transportation networks, metablolic pathways, and computer networks.
展开▼