Using Machine Learning to Build Models of Relativistic Particle Acceleration Relevant to High Gain ICF (Plasma Strand Project)

Above image: Automated EPOCH particle-in-cell simulations provide training data for a machine learning algorithm.  In this case this algorithm determines the efficiency of x-ray production by bremsstrahlung as a function of target ion number density (n_i) and thickness (d), as well as the laser intensity (I_0),  Taken from Smith et al.arXiv:2411.02079

Supervisors: Chris Ridgers (University of York), Jiannan Yang (University of York) and Thiyagalingam Robinson (Ada Lovelace Centre).

High intensity lasers create plasmas containing strong fields. In these fields electrons and
ions are accelerated to relativistic energies over very short distances. This rapid
acceleration causes them to emit x-rays and gamma-rays. These energetic particles and
radiation are useful for Inertial Confinement Fusion (ICF) in several ways. A promising
high gain ICF scheme – fast ignition – relies on heating compressed fusion fuel with
relativistic electrons or ions. Energetic x-rays can be used to probe compressed ICF
pellets. Both of these are highly promising avenues of research for moving towards high
gain ICF, essential for future power generation based on inertial fusion. Fast ignition is
potentially a much higher gain scheme than standard ignition as the fuel can be
compressed relatively cold (and is therefore easier to compress). Advanced diagnostics
on current ICF experiments (for example x-ray radiography) can help us determine the
causes for loss target performance, for example by hydrodynamic instabilities and the
reasons for high performing targets, such as the ignition shot 210808 on the National
Ignition Facility. Such an understanding will help the design of targets for higher gain.

This project will focus on the development of surrogate models for particle acceleration
and radiation generation in laser-plasma interactions. Such models will be necessary if
we are to realise fast ignition as we will need to understand how to tune the electron or ion
sources in order to minimise the laser energy required and so maximise the gain. We also
need to be able to control our laser based x-ray sources if we are to apply them to
radiography of ICF targets. High intensity laser plasma interactions are typically modelled
using particle in cell (PIC) codes. These codes are computationally very expensive to run,
requiring the use of large supercomputers. The particle acceleration often depends on
nonlinear processes in the plasma. This means we need to do lots of simulations over a
wide range of parameters to build models of the acceleration. We have devised a new
methodology for doing this using machine learning. The expensive PIC simulations are
used as training data for a Gaussian Process Regression (GPR) algorithm. GPR can
work well with sparse data and we apply active learning to put the simulations where they
are required in parameter space to minimise the error in our surrogate model. This new
approach has yet to be applied to particle acceleration and radiation emission. That will be
the focus of this project.

This project will be primarily computational, applying GPR to build surrogate models of
PIC simulation data. Interaction with experimentalists will be encouraged as these models
are highly useful as they can be run on the fly and can help steer experiments. This will
be especially important as we move into the era of high repetition rate lasers at the
Central Laser Facility (Rutherford Appleton Laboratory) with the construction of the new
EPAC facility. This project is a collaboration between the University of York, The Ada
Lovelace Centre and the Rutherford Appleton Laboratory. You will gain skills in high
performance computing and machine learning as well as in how to apply these to practical
scenarios realisable in the laboratory.

We will collaborate with the Central Laser Faclity and Ada Lovelace Centre. There is opportunity to travel to these as well as to international conferences.

This project is offered by University of York. For further information please contact: Christopher Ridgers (christopher.ridgers@york.ac.uk).

This project may be compatible with part time study, please contact the project supervisors if you are interested in exploring this.

For details on how to apply, please visit: Apply