Fast cosmological simulations
Supervisor: Oliver Hahn
Co-supervisor: Florian List
Contact information: oliver.hahn@univie.ac.at
Expected duration: 9 months
Project description & Goals:
The quest for physics beyond the standard models of cosmology and particle physics is the primary motivation for upcoming multi-billion euro experiments and observatories in space and on earth (Euclid, LSST, DESI,...). They aim to answer: what is the nature of dark matter/dark energy, the mass of neutrinos, and what can we learn about the earliest inflationary phase of our Universe by observing galaxies and the intergalactic medium? Exploiting the data from these surveys requires extremely accurate and fast modelling of the matter distribution in the Universe with the ultimate aim to integrate simulation codes into inference pipelines. In Vienna, we are developing a new framework to predict cosmological observables (such as the mass distribution, the Lyman-alpha forest, the distribution of galaxy clusters) that leverages several new technologies. In the context of this project, there are multiple possible master projects available (ranging from adding new physical component, JAX-GPU code development, over the implementation of a hydro solver, the integration of astrophysical models, to the modelling of astrophysical observables).
Working plan & Milestones (including final thesis):
Depending on the specific project, the detailed plan may deviate somewhat:
- Literature study focused on fast forward models and large-scale structure cosmology
- Develop and implement one new aspect of the model inside of the larger project (e.g. porting to multi-GPU, adding fast hydro solver, modelling astrophysical processes, modelling astrophysical observables)
- Validate the model/implementation and compare the performance against standard high-resolution simulations, document the impact of parameter choices on the result
- Write thesis, and depending on progress, a short paper presenting the results to the community
Requirements / special skills:
Good coding skills in Python, some familiarity with cosmology/large-scale structure beneficial, should have followed the numerics lab course
References:
(note that some of these are very technical, much beyond what we would expect a beginning student to understand)
Feng et al — FastPM: a new scheme for fast simulations of dark matter and halos [first paper on fast integrators] https://arxiv.org/abs/1603.00476
List & Hahn -- Perturbation-theory informed integrators for cosmological simulations [new integrators used in our framework] https://arxiv.org/abs/2301.09655
Modi, Lanusse & Seljak -- FlowPM: Distributed TensorFlow Implementation of the FastPM Cosmological N-body Solver [similar framework] https://arxiv.org/abs/2010.11847
Angulo & Hahn -- Large-scale dark matter simulations [general review] https://arxiv.org/abs/2112.05165