CAMB: Unphysical low ell CBB with high accuracy settings

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply
Pavel Motloch
Posts: 13
Joined: October 14 2016
Affiliation: CITA

CAMB: Unphysical low ell CBB with high accuracy settings

Post by Pavel Motloch » October 14 2016

Reporting a numerical issue, using May16 version of CAMB. Full parameter file below.

Lensed C_L^BB are evaluated as negative for the lowest few L. Issue can be tracked down to evaluation of d_2m2 in lensing.f90/CorrFuncFullSkyImpl for low ell and small theta.

When you plot d_2m2(ell = 2) as a function of theta, you see numerical noise kick in below theta ~ 0.01 radian and increase with theta -> 0. This then leads to unphysical C_L^BB. Presumably important only for very high accuracy setting, when you evaluate integrand at very small theta. Might be possible to remove by a switch which applies Taylor expansion at lowest angles.

Parameter file:
output_root = fiducial_precise/fiducial_high_ell
get_scalar_cls = T
get_vector_cls = F
get_tensor_cls = F
CMB_outputscale = 7.42835025e12
get_transfer = F
accuracy_boost = 4.1
l_accuracy_boost = 3.1
high_accuracy_default = T
do_nonlinear = 3
l_max_scalar = 20000
k_eta_max_scalar = 100000
do_lensing = T
lensing_method = 1
w = -1.000000
cs2_lam = 1
hubble = 67.5
use_physical = T
ombh2 = 0.0222
omch2 = 0.1197
omnuh2 = 0.000645144
omk = 0.000000
temp_cmb = 2.7255
helium_fraction = 0.25
massless_neutrinos = 2.046
nu_mass_eigenstates = 1
massive_neutrinos = 1
share_delta_neff = T
nu_mass_fractions = 1
transfer_high_precision = T
halofit_version = 4
DebugParam = 0.000000000000000E+000
Alens = 1.00000000000000
reionization = T
re_use_optical_depth = T
re_optical_depth = 0.06
re_delta_redshift = 0.5
re_ionization_frac = -1
re_helium_redshift = 3.50000000000000
re_helium_delta_redshift = 0.500000000000000
re_helium_redshiftstart = 5.00000000000000
pivot_scalar = 0.050000
pivot_tensor = 0.050000
initial_power_num = 1
scalar_spectral_index(1) = 0.9655
scalar_nrun(1) = 0
scalar_nrunrun(1) = 0.000000000000000E+000
scalar_amp(1) = 2.196e-09
RECFAST_fudge_He = 0.86
RECFAST_Heswitch = 6
RECFAST_Hswitch = T
RECFAST_fudge = 1.14
AGauss1 = -0.140000000000000
AGauss2 = 7.900000000000000E-002
zGauss1 = 7.28000000000000
zGauss2 = 6.73000000000000
wGauss1 = 0.180000000000000
wGauss2 = 0.330000000000000
do_lensing_bispectrum = F
do_primordial_bispectrum = F
initial_condition = 1
scalar_output_file = scalCls.dat
lensed_output_file = lensedCls.dat
lens_potential_output_file = lenspotentialCls.dat
accurate_polarization = T
accurate_reionization = T
accurate_BB = T
derived_parameters = T
version_check = May16
do_late_rad_truncation = T
feedback_level = 1
output_file_headers = T
massive_nu_approx = 1
number_of_threads = 0
use_spline_template = T
l_sample_boost = 3.0

Post Reply