Chemical Crossover

During chemical reactions, molecules interact to produce new molecules. Some of the reaction mechanisms are very close to the way gene combine to produce new genes. Like the classical genetic crossover operator, chemical crossover occurs when two molecules exchange the atomic material they are composed of. Molecules do so in order to produce new and more stable molecules. Several differences exist however between the genetic crossover and its chemical version. Molecules are not coded as binary string but as computational tree instead. Moreover this computational tree, due to the symmetry in the way atoms are connected, must be organized in a strictly ordered way. Following a crossover, the resulting molecules must be reshaped according to precise organization rules. The crossover involves the exchange of single or multiple links. The fitness is distributed through the molecule such that only " better " molecules can result from the crossover. In this paper the chemical crossover and the computer simulation will be discussed within several perspectives: chemical, Alife and engineering. Simulation results will be presented for a simple chemical reactor composed of four atoms with different valence, and allowing molecules to deterministically or randomly interact according to the single-link-crossover. 1 INTRODUCTION From its origins, Alife has always taken as fundamental the conception of software environments able to model the appearance and the disappearance of complexes constructed from simpler complexes or basic elements. This preoccupation leads to a succession of keystones in the Alife production such as Fontana's alchemy (1992), Kauffman's autocatalytic networks (1993) or Holland's echo models (1995). These software environments, like the major part of work being done (or which should be done) in the " Alife spirit " , could have three major applications. Following an adequate parameterization, they could be used by chemical or biological practitioners who could find in these " computational platform " a helpful way to model a particular system. We can then speak of a set of computational design patterns, or software shell, which should easily be adapted to the simulation of a particular chemical or biological environment. On the other hand, these simulations can stand on their own and allow the discovery of generic laws, expressed in mathematics or linguistic terms, characterizing the behaviour of emergent and complex systems. For instance, in Kauffman's simulation of Boolean nets (1993), some laws establish the stability or instability of generic networks as a function of the range of …