Artificial Chemistries have, up to now, focused on building micelles, [Mayer and Rasmussen, 1998], believed to be the starting point to generate cell boundaries, or in building self-maintaining sets in non spatial environments.
We try to overcome the bridge between these two modus operandi by presenting a model of artificial chemistry embedded in a spatial structure.
To do this we rebuild Fontana’s model ‘AlChemy’ with some differences. In particular we change the operation to be more realistic. Now, as elements react, they are effectively used up in the reaction, but the number of released elements is not fixed anymore, but dependent on the particular reaction.
We also embed the whole system onto a graph, in particular a complete planar graph. Such a structure has never been used previously (as far as we know) as a base for artificial life models, but proves to be absolutely ideal, giving simplicity side by side with flexibility. In fact it is very similar to a normal Cellular Automata, but gives the possibility to new nodes (elements) to appear anywhere in the experimental space.
The model proved to be able to support and generate self-maintaining infinite sets (akin to L1 organizations in AlChemy). These sets tend to organise themselves in bubbles of activity surrounded by elements connected through fixed links. Those elements (and links) being similar to boundaries, divide the bubbles of activity, and as new elements get generated inside the bubble, they tend to either make the boundary thicker, or split the bubble into two parts, with a process similar to cell reproduction. The same self-maintaining set that was generated one of the bubbles will now generate both the two new copies. We present this as a possible starting point to generate increasingly complex systems.