The problem for evaluation was given to us by the WarpX team. Physically, as depicted in the picture above, the problem is the focusing of high-order harmonic beams by a relativistic plasma mirror. Such plasma mirrors can be formed by focusing a femtosecond laser pulse of intensity I > 1018; W/cm2; with high temporal contrast on a solid target such as plastic or glass. Upon reflection on the relativistically oscillating mirror surface, high-order harmonic beams are generated by the Doppler effect in the reflected field propagating along the specular direction. For Gaussian laser profiles at focus, the inhomogeneous laser radiation pressure can curve the plasma mirror surface, forming a parabolic plasma mirror, which can in turn tightly focus the harmonic beams. This process can lead to extreme intensities I > 1025 W/cm2 at plasma mirror focus, and is currently a very promising candidate to access yet unobserved quantum electrodynamic processes such as electron-positron cascades or vacuum breakdown. For our performance testing, the laser is omitted in order to simplify the problem, because it does not take much computational time.
The domain of the problem we ran on WarpX and PICSARlite is the rectangle from points (-1.42e-5, -8.5e-6, 0) to (2.8e-6, 8.5e-6, 3.4e-5). The plasma consists of two species of particles, electrons and protons, both distributed in the rectangular subregion from (0, -8.4e-6, 1e-6) to (2.5e-6, 8.4e-6, 3.3e-5), with 2 particles of each species per grid cell. The particle-containing region has dimensions 2.5e-6 × 1.68e-5 × 3.2e-5, within the full domain of dimensions 1.7e-5 × 1.7e-5 × 3.4e-5. This is 14.7%, 98.9%, and 94.1% of the length of the domain in the three dimensions.
We did not do any validation to check that WarpX and PICSARlite give the same answers. We asked the WarpX team for ways to validate results, but they told us that the different output formats of the two programs as well as the nature of the output itself make a comparison of results difficult.