Particle-beam technology has wide applications in science and industry. Specifically, high-energy x-ray production is being investigated for FLASH radiotherapy, 14 MeV neutrons are being produced for fusion energy production, and compact electron accelerators are being built for medical-device sterilisation. In each instance it is critical to guarantee that the particle beam is delivered to the end user with the correct makeup, and also to ensure that secondary particles created from scattering interactions are shielded from technicians and sensitive equipment. There is no precise way to predict the random walk of any individual particle as it encounters materials and alloys of different shapes within a complicated apparatus. Monte Carlo methods simulate the random paths of many millions of independent particles, revealing the tendencies of these particles in aggregate. Assessing shielding effectiveness is particularly challenging computationally, as the very nature of shielding means simulations produce low particle rate.
A common technique for shielding calculations takes these random walk simulations a step further by applying variance reduction techniques. Variance reduction techniques are a way of introducing biases in the simulation in a smart way to increase the number of particles emerging from the shielding, while still staying true to the total conservation of matter. In some regions within the shielding, particles are split into independent “daughter” particles with independent pathways but some common history. They are given a weight value, so the overall flux of particles is kept constant. In this way, it is possible to predict the behaviour of a one-in-a-million event without having to simulate one million particle trajectories. The performance of these techniques is shown in figure 2.
These kinds of simulations take on new importance with the global race to develop fusion reactors for energy production. Materials will be exposed to conditions they’ve never seen before, mere feet from the fusion reactions that sustain stars. It is imperative to understand the neutron flux from fusion reactions and how they affect critical components in the sustained operation of fusion facilities if they are going to operate to meet our ever-growing energy needs. Monte Carlo simulation packages are capable of both distributed memory (MPI) and shared memory (OpenMP) parallel computation on the world’s largest supercomputers, engaging hundreds of thousands of cores at once. This enables simulations of billions of particle histories. Together with variance reduction, these powerful simulation tools enable precise estimation of particle fluxes in even the most deeply shielded regions.
RadiaSoft offers browser-based modelling of neutron radiation transport with parallel computation and variance reduction capabilities running on Sirepo, their browser-based interface. Examples of fusion tokamak simulations can be seen above. RadiaSoft is also available for comprehensive consultation in x-ray production, radiation shielding and dose-delivery simulations across a wide range of applications.
