Data Challenge Map Sets 04.YY

From CMB-S4 wiki
Revision as of 10:05, 21 February 2018 by Pryke (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigationJump to search

Feb 6 2018, Clem Pryke

We have made sets of simulated CMB-S4 fsky 3% maps with bands, resolutions and noise levels as per the CDT report "Science Traceability Matrix" - these are dubbed "experiment definition 04. The specs are repeated on the Experiment Definitions page.

These are combined with the seven foreground models as used in appendix A of the CDT report to form map sets 04.00 through 04.06. What these foreground model are is noted on the Sky Models page. These sets of maps are available on NERSC in directory /project/projectdirs/cmbs4/data_04.00 etc. The intent is have multiple groups or individuals re-analyze these maps to determine the precision with which the tensor-to-scalar ratio can be recovered with a particular algorithm for the given noise levels and foreground behavior.

Here is a plot showing the Q maps for realization 0001 and its components where each column is a frequency band. All the maps are filtered to the appropriate resolution of the given band, and also have ell<30 filtered out to simulate the inability of ground based experiments to recover the largest angular scales. The top row shows the lensed-LCDM input. The second row is tensors with the color stretch zoomed in by factor 20 (for r=0.01). The third row is the foreground model - in this example one of the PySM variants. The fourth row is the noise component which blows up towards the edge of the field in a way which is typical of small aperture observations where the instantaneous field of view is large compared to the field size. The noise is 1/f+white with the specs as per the "Experiment Definitions" page linked above. The bottom row is what one actually gets in a real experiment - all of the above added together. Note that a pdf version is available which allows one to zoom in and see the full resolution of the maps. (Sorry for the goofy cylindrical projection - the sky patch is actually round on the sky, centered on RA=0, Dec=-45deg.)

180206 f1.gifMedia:180206_f1.pdf

The plot below shows the seven foreground models in each of the bands. We see that they all roughly agree as to the level of synchrotron but differ strongly in the level of dust emission. The three PySM models are based on the same templates and differences between them are subtle. The Flauger/Hensley MHD model shows how the sim box has been manipulated to cover the desired obs region.

180206 f2.gifMedia:180206_f2.pdf

For a realistic "data challenge" re-analysis one would use only the "noise" and "combined" maps as shown above plus the following: The table giving the band centers/widths and resolutions etc is at /project/projectdirs/cmbs4/expt_xx/04/params.dat and the relative hits map (i.e. inverse noise variance apodization mask) is at /project/projectdirs/cmbs4/expt_xx/04/rhits/n0512.fits. In addition it may be legitimate to assume that one knows the lensed-LCDM spectrum already (from studies of larger sky regions) and this is at /project/projectdirs/cmb/data/generic/cmb/ffp10/cls/.

One cheat which is going on here is the treatment of de-lensing. In principle this should be done by reconstructing the lensing using higher resolution maps - and sim maps suitable for doing that are coming soon. In the meantime for this round we provide combined maps where the lensing component has been artificially scaled down to some fraction of reality. For instance file cmbs4_04p06_comb_AL0p1_f020_b11_ellmin30_map_0512_mc_0000.fits assumes factor 10 delensing.