Data Challenge Maps for Commander

Apr 21 2021, Clem Pryke

In Data challenge maps I I introduced some simulated maps to be used for reanalysis to help determine PICO performance. These have been used successfully by Mathieu and others.

The generation machinery was borrowed from CMB-S4 work where it was used at nside=512 for small aperture telescopes (and max frequency 270GHz). When I made the PICO maps nside=512 was kept but the beam sizes were taken at face value (arcminute at the highest frequencies).

The LCDM and noise components were set to cutoff at 2 times nside. However the PySM foreground maps included small scale fill in out to 3 times nside. This meant that for the smaller beam sizes from 2 to 3 times nside the foreground model was the only active component. Apparently the Commander algorithm requires that the maps be noise dominated at the highest multipoles so this caused problems.

To try and address this I have made some new 91.XX maps with 3 changes versus the existing 90.XX ones:

• A lower limit of 20 arcminutes is placed on the beam sizes (affecting 51.8GHz and above)
• Noise a_lm's out to 4 times nside are included.
• A cutoff at 2 times nside has been imposed the foreground component (for consistency with the LCDM component).

The spectra below illustrate for the 30, 108 and 321 GHz bands - “Set 90” is old and “Set 91” is new. The 20 arcmin beam roll-down hard before the ell=1024 cutoff and the noise extends out to ell=1536 (the max ell plotted). If these maps look like they will be acceptable I will put them on NERSC.