When: Wednesday, December 7, 2022, 3:30 PM - 5:00 PMWhere: Physics (G. O. Jones building) Room 516 & online, Mile End
Speaker: Bartolomeo Fiorini (QMUL)
The large-scale structure (LSS) of the universe that we observe in galaxy surveys is believed to originate from the gravitational evolution of tiny density fluctuations present in the early universe and observed in the CMB. The gravity model governing this evolution is not yet well constrained but with the large volume probed by the next generation of galaxy surveys we will have the chance to put tight constraints on the gravity model on cosmological scales. However, to achieve this goal theory must keep pace with observations, aiming at percent level accuracy deep into the non-linear regime of structure formation. Cosmological simulations are an essential tool to reach this accuracy but their computational cost is very high. Fast approximate simulations methods can reduce the cost of theory predictions and together with emulators are a promising tool to enable cosmological inference with the non-linear regime. In this talk I will describe how these approximate methods can be applied to MG theories to simulate the dark matter distribution and introduce a pipeline for the efficient production of mock galaxy catalogues in MG. Applying the pipeline to 2 MG theories and the vanilla GR model, I will show what traces of MG we can expect to see in LSS probes like RSD, bispectrum and voids. To showcase the power of this method in combination with machine learning techniques, I will present an emulator for the matter power spectrum in a popular MG model (nDGP) which reduces the cost of theoretical predictions from millions of CPU hours to milliseconds.