This page collects a sample of 6000 x 6000 pixels mock radio observations used to train our Cosmodeep Convolutional Neural Network (Gheller et al. 2018). 

COSMODEEP
TRAINING
RADIO FIELDS

We recently produced ​ very big cosmological MHD simulations to study the evolution of magnetic fields in large-scale structures and quantify the observable signal from the shocked cosmic web in synchrotron.
The ​ sky models generated in this project can be used to test which strategies are best to lead to a detection of the radio cosmic web, with existing (e.g. JVLA, LOFAR, ASKAP) or incoming radio telescopes (most noticeably the SKA).
All simulations have been produced on ​​​ Piz-Daint at CSCS
in 2014, using the ENZO code.
Our data are share using the EUDAT service.
Here below it is possible to download our fits files containing the sky models generated with our MHD runs. They are meant to represent large areas of the sky and can be used for testing of survey strategies with radio telescopes.

They have been specifically designed to mimic ASKAP observations in simplistic way, including the effects of the 10" resolution beam, thermal noise of 10mJy/beam.

Each image is a 2000 x 2000 pixels field, every pixel=10". The intensity scale of every pixel is given in [μJy/arcsec^2]
Sky and noise models
​61 images ~ 16 Mb each
 
Radio model
  • Askap (EMU)-like observations
  • beam(=pixel)=10"
  • rms noise=10 μJy/beam
  • FOV=2000 x 2000 pixels
  • no radio artefacts, radio galaxies, foregrounds
 
Simulation details
  • stacked volumes up to z=0.5
  • primordial seeding B=0.1nG
  • spatial resolution dx=40kpc/cell
  • synchrotron emission from cosmic shocks
 
GOTO B2share REPOSITORY
CONTRIBUTORS:
C. Gheller (EPFL, Lousanne, Switzerland)
F. Vazza (Università di Bologna)
A. Bonafede (Università di Bologna)