Skip to main content

Cosmological joint-probe analyses: constraining the effect of baryon physics on the matter distribution

Periodic Reporting for period 1 - Joint-probe analyses (Cosmological joint-probe analyses: constraining the effect of baryon physics on the matter distribution)

Reporting period: 2018-04-02 to 2020-04-01

Among the most pressing questions in physics are the nature of dark energy and dark matter. The former is the driving force behind the current accelerated expansion of the Universe while the latter accounts for the vast majority of matter and is responsible for the growth of the large-scale structure in the Universe. No satisfactory explanation has been put forward for either of these two mysterious substances, however.

Future cosmological surveys, such as Euclid, will elucidate the nature of dark matter and dark energy by measuring the distribution of matter and the growth of structure in the Universe to unprecedented accuracy. These high-precision measurements put stringent requirements on the accuracy of the modelling and statistical tools. However, the highly non-linear processes that govern the distribution of matter at small scales pose formidable challenges to modelling and statistical analysis, since at these small scales poorly understood effects of galaxy formation on the matter distribution become important.

The first objective of the JOINT-PROBE ANALYSES action was to constrain the effects of galaxy formation on the matter distribution by performing a joint-probe analysis of the three cross-correlations between weak gravitational lensing from galaxies, lensing of the cosmic microwave background (CMB), and the thermal Sunyaev-Zeldovich (tSZ) effect.

The second objective aimed at providing new tools to estimate the statistical uncertainty in cosmological surveys, specifically using resampling methods based on recent advances in the statistical literature.

Both objectives aim at enabling Euclid, and other future large-scale cosmology experiments worth hundreds of millions of Euros, to reach their full potential.
An important observational systematic are contributions by the cosmic infrared background (CIB) to the observed tSZ signal.
The CIB traces the same large-scale structures of matter in the Universe as does the tSZ effect and gravitational lensing. This makes disentangling these signals difficult. The Fellow quantified the effect of CIB contamination on the observed lensing-tSZ cross-correlation (Yan et al. 2019, arXiv:1809.09636).

In order to model the expected cross-correlation signal between gravitational lensing and the tSZ effect, the Fellow developed a new modelling framework that jointly describes the distribution of dark matter, gas, and stars in the Universe (Mead, Tröster et al. 2020, arXiv:2005.00009).

This model allowed for a detailed, quantitive forecasting of the constraining power of the planned joint analysis of cross-correlations of the gravitational lensing of galaxies, the tSZ effect, and lensing of the CMB. This forecasting analysis revealed that the envisaged joint analysis was suboptimal, since the cross-correlation between galaxy and CMB lensing contributes negligibly and the CMB lensing-tSZ cross-correlation is beset by observational systematics. On the other hand, it was found that joint analysis of cosmic shear, and the cross-correlation between weak lensing of galaxies and the tSZ has the potential to strongly constrain both cosmological parameters and the effects of galaxy formation, as set out in the objectives of the action.

To aid the statistical inference of the tSZ-lensing cross-correlation analysis, the Fellow developed a method to augment dark-matter only simulations with gas using deep learning techniques (Tröster et al. 2019, arXiv:1903.12173). This important development allows the rapid generation of mock data that would otherwise require running computationally very expensive hydrodynamical simulations that incorporate galaxy formation.

In preparation of the joint analysis of cosmic shear and the tSZ-lensing cross-correlation, the Fellow joined the Kilo-Degree Survey team to gain expertise in the analysis of cosmic shear data and develop the required tools, resulting in a number of high-impact publications (among others: Tröster et al. 2020, arXiv:1909.11006; Asgari, Tröster et al. 2020, arXiv:1910.05336; Hildebrandt et al. 2020, arXiv:1812.06076).

Resampling techniques, such as the bootstrap, are an attractive statistical method to find the properties of observed data, such as their statistical uncertainty. To validate the performance of these methods they can be inspected in simplified situations where the exact result is know. For the case of two-point correlations, such as the proposed lensing-tSZ cross-correlation, it was found that no such exact result exists, neither in the astronomy nor statistics literature. The Fellow as now derived this exact result (Tröster, 2020, in prep.).

The results of the action were disseminated in international conferences and workshops (nine in total, including invited and contributed talks), as well as, scientific publications. At the time of this report, the actions has produced 15 publications, of which 13 have been peer-reviewed.

No website has been developed for the project.
The different projects of this action of contributed to pushing the state of the art in the field forward. Notably the work (Tröster et al. 2019, arXiv:1903.12173) on using deep learning techniques to emulate the output of computationally expensive hydrodynamical simulations based on much cheaper dark-matter-only simulations.

The analysis of clustering of galaxies, as well as their joint analysis with weak lensing (Tröster et al. 2020, arXiv:1909.11006) revealed a new approach to extracting information from these data that had been overlooked in the field for years.
Dark matter density (left), true pressure (middle), and pressure predicted by deep learning model.