Simulations are often required to create successful designs and interpret experimental results. Simulations of plasmas have many applications including medicine (therapy and diagnostics), engineering (etching and cleaning) and power (fusion generators and dielectric insulators). All of these simulations rely on accurate and detailed inputs and the most crucial of these are collision frequencies for scattering processes.
A collision frequency can be trivially obtained from the cross section for a process. Unfortunately, calculating or measuring individual cross sections is difficult. An alternative is to perform a “swarm experiment” in which a measurement is made of a group of electrons or positrons drifting through a gas or liquid. These measurements are sensitive to the underlying processes, such as elastic scattering, ionisation, attachment and annihilation. Hence, transport measurements such as drift velocity and diffusion provide an indirect probe for the microscopic properties of the system.
Normally simulations take microscopic cross sections as their inputs and give transport properties as their output. The reverse process is known as the “inverse swarm problem”. We aim to extract the microscopic cross sections from measured transport properties.
This project will make use of neural networks to train an analysis program to extract an “effective cross section set” to model a transport system. The network will then be used to analyse transport measurements in biomolecules which are reasonable substitutes for DNA.