lenspix NaN in output Q and U map when lmax > 7000

Use of Healpix, camb, CLASS, cosmomc, compilers, etc.
Post Reply
Zhen Hou
Posts: 6
Joined: March 02 2009
Affiliation: KICP, University of Chicago

lenspix NaN in output Q and U map when lmax > 7000

Post by Zhen Hou » October 16 2014

Dear Antony,

I was trying to run lenspix with polarization turned on for high nside values, e.g., nside=4096 and 8192. However, I find all the elements of the output Q and U map are '-NaN' if I set lmax larger than 7000. In fact, for lmax lower than 7000, I only tested lmax=6500 that the output looked fine. The T map is always good for large lmax.

I have checked the input Cls which contained ell from 2 to 15100 with delta_ell=1 for [2,10000] and delta_ell=100 for [10000, 15100]. I don't think there is problem in the input Cls.

I am using lenspix version April 2014 and Healpix 3.11, both compiled by intel fortran 14.0.2.

Here are relevant lines in my params.ini

nside = 4096
lmax = 7000
cls_file = planck_lensing_wp_highL_bestFit_20130627_lenspotentialCls.dat
lens_method = 1
interp_factor = 0.375
interp_method = 1
mpi_division_method = 3
want_pol = T
want_map = T
want_alm = F
rand_seed = 320

I am not sure if any one met this problem before. Could you help to look into it? Thank you!


Pavel Motloch
Posts: 6
Joined: October 14 2016
Affiliation: CITA

lenspix NaN in output Q and U map when lmax > 7000

Post by Pavel Motloch » March 16 2017

Hi Zhen,

my guess is your input Cls are flawed as they do not satisfy a physicality condition C_TT*C_EE > C_TE^2. Then Lenspix generates NaN a_lm, which then propagate into the maps.

This can happen if in CAMB you choose scalar lmax > 8000 (which is the lmax of the provided high precision interpolation file) due to complicated CAMB interpolation scheme.

Can hack this in Lenspix code which generates alms, *_Sim or so.


Post Reply