Simulating realistic laser wakefield acceleration scenarios on the fly using GPUs

In order to optimise internal and external injection of electrons into laser-driven wakefields generated and the subsequent acceleration it is necessary to get fast response from simulations. This means that when considering a number of parameters such as plasma density, laser intensity, laser focal position and size or plasma density gradients, the optimum set of parameters is not easy to obtain by realistic particle-in-cell (PIC) simulations.
PIC simulations take into account the full dynamics of the particles in the interaction of the laser pulse with the plasma and thus give a detailed and realistic view of the plasma dynamics that can be used for optimisation. However, such detailed simulations usually require a lot of computing power not easily available to small-scale laboratories and universities.
We present PIConGPU, a fully relativistic, scalable implementation of the PIC algorithm on graphic cards (GPUs) that gives a speed up of orders of magnitude compared to PIC implementations on CPUs. Our code allows for an execution time below one nanoseond per macroparticle simulated (2D3V). This gives the user “real-time” feedback from the simulation, seeing on the fly the influence of the chosen parameter set on the injection and acceleration of electrons into the laser-driven wakefield. With this tool, the optimisation of injection schemes for the laser wakefield acceleration of electrons using realistic particle-in-cell simulations is in reach.