Explaining the dark matter mystery

The role of Anders Kvellestad at the University of Oslo and his fellow researchers’ GAMBIT project is to test such new theories against results from all types of particle physics experiments. This includes results from the Large Hadron Collider at CERN and from many different experiments that try to detect dark matter particles.

Every new theory comes with a set of free parameters – essentially knobs that can be tuned to alter the theory’s predictions. The GAMBIT code combines smart sampling algorithms, parallelisation, and optimised physics simulations to explore the parameter space of new theories and test the predictions against experiment results. This way, the researchers identify which theories can be ruled out and which should be investigated further in future experiments. They were recently awarded computing time on the PRACE Tier-0 systems.

A perhaps distinct feature of the GAMBIT project is that the research programme is divided into multiple independent subprojects, each investigating a different particle physics theory. This modularity has made it easier to ensure good use of the allocated computing resources, even in cases where one subproject encounters an unexpected delay.

The GAMBIT project was awarded its first PRACE allocation, a three-year Project Access grant, back in 2017. In preparation for that call, Kvellestad and his fellow researchers applied for and were awarded PRACE Preparatory Access to carry out performance tests of the GAMBIT code on the relevant HPC system. This was critical for the success of their first application. 

This time around, they thus had the benefit of building their application on three years of experience with running GAMBIT on different PRACE HPC systems, which is the computational backbone of the project. A project that has grown to involve around 70 researchers worldwide. 

The new award will enable the researchers to extend their programme further and to investigate new particle physics theories that have never been subjected to the type of large-scale and robust statistical assessment that can be performed with GAMBIT.