Evolution in Materio: Exploiting the Physics of Materials for Computation

TL;DR

This paper explores using evolutionary algorithms to program various physical substrates, such as CMOS, liquid crystals, and quantum dots, for specialized computation by leveraging their intrinsic physical interactions.

Contribution

It introduces multiple novel methods for programming diverse physical materials for computation through evolutionary algorithms, demonstrating their potential across different scales and physical phenomena.

Findings

Successful evolution of a CMOS oscillator

Development of a liquid crystal tone discriminator

Control of resistance in charge density wave systems

Abstract

We describe several techniques for using bulk matter for special purpose computation. In each case it is necessary to use an evolutionary algorithm to program the substrate on which the computation is to take place. In addition, the computation comes about as a result of nearest neighbour interactions at the nano- micro- and meso-scale. In our first example we describe evolving a saw-tooth oscillator in a CMOS substrate. In the second example we demonstrate the evolution of a tone discriminator by exploiting the physics of liquid crystals. In the third example we outline using a simulated magnetic quantum dot array and an evolutionary algorithm to develop a pattern matching circuit. Another example we describe exploits the micro-scale physics of charge density waves in crystal lattices. We show that vastly different resistance values can be achieved and controlled in local regions to essentially construct a programmable array of coupled micro-scale quasiperiodic oscillators. Lastly we show an example where evolutionary algorithms could be used to control density modulations, and therefore refractive index modulations, in a fluid for optical computing.