An Investigation of Underlying Physical Properties Exploited by Evolution in Nanotubes Materials

Computational materials, e.g. Single-wall carbon nanotubes and polymer nanocomposites, have been evolved to solve complex computational problems. Such blobs of material have been treated as a black box, e.g. Some input is encoded, some configuration signals are evolved to "program" the material machine, and some output is decoded. However, how the computation is performed, i.e. Which physical properties are exploited by evolution to solve a given computational task, is not well understood. The general idea is that some undelying physical properties of the chosen material are exploited, e.g. Capacitance, resistance, voltage potential, signal frequency, etc. In this paper we investigate which practical strategies are exploited by evolution on a simple (non-abstract) task: maximize or minimize amplitudes of output signals when square waves are used as input. This allows identifying an evolvability range for materials with different physical characteristics, e.g. Nanotubes concentration. Inspection of evolved solutions shows that the strategies used by evolution to exploit physical properties are often unanticipated. This work is done within the European Project NASCENCE.