EPFL/CERN/UNIGE Discussion Sessions

Wednesday 21 Sept 2016, 10:30 → 12:30 Europe/Zurich

4/2-037 - TH meeting room (CERN)

To prompt collaboration among the different groups in cosmology in the Léman area, we hold a common discussion session every month and a half. In those, recent news in the field are addressed by experts at a level that we can all understand. Please, suggest papers/subjects to the local organizers. The suggested papers for this session are:

 

 

 

10:30 → 11:15 H0LiCOW V. New COSMOGRAIL time delays of HE0435-1223: H0 to 3.8% precision from strong lensing in a flat ΛCDM model

http://arxiv.org/abs/1607.01790

Speaker: Vivien Bonvin

11:30 → 12:15 The Gaussian streaming model and Lagrangian effective field theory

Authors: Zvonimir Vlah, Emanuele Castorina, Martin White
Journal Club Discussion

Speaker: Matteo Biagetti

11:32 → 11:33 Cancelled

I will discuss how the pkdgrav3 cosmology code managed to complete the world's largest dark matter simulation to date on the Piz Daint supercomputer at CSCS (Lugano, Switzerland). Pkdgrav3 development was motivated by the challenge of using all of this system's resources optimally, the RAM, CPU, GPU, Caches, Network and Disk. Dealing with the last constraint, the finite amount of disk space, meant that we gave up on some conceivable post-process analysis. Rather, we now view the simulation code more like a telescope, to which one attaches new future instruments (on-the-fly analysis code) and reruns the simulation when needed. The "instruments" we had time to implement are yet few and crude, but much of the development in the next years will aim to improve this situation. I will try to show some nice images from the simulation, to give a sense of both the large scale and the high resolution it achieves. We are currently processing the data with the final goal of having a mock galaxy catalogue for Euclid as well as weak lensing maps. This simulation, and future refinements of it are critical to designing and testing the pipeline of Euclid as well as ultimately extracting parameters from the data. It is a difficult task to achieve the required precision in the theory of non-linear structure formation at the level deemed necessary, even where simulations such as these are used as "control points". Effects from radiation, neutrinos and most importantly baryons are currently not handled self-consistently at the level which is possible with such pure dark-matter simulations. This challenge of achieving precision theory in cosmology will remain for simulators in the coming years and I will look ahead and speculate about a roadmap we may follow.

Speaker: Prof. Joachim Stadel (University of Zürich)