CAMB_sources weird kink in lensing spectrum at l~600

[kiyo masui]

Hi all,

I'm using camb_sources to compute the weak lensing spectrum. I have changed the Window\_f\_a function to a smoothed step:

Code: Select all

! Smotthed window function
k = 100.0
winamp =1+exp(-2*k*((1/a-1)-Win%Redshift+Win%sigma))
winamp =winamp+exp(2*k*((1/a-1)-Win%Redshift-Win%sigma))
Window_f_a=1.0/winamp/2/Win%sigma/a**2

I'm getting a weird kink in the lensing spectrum, which is most prominent if I turn on Halofit, see the plot below.



The solid blue curve is approximate and calculated in a way that has nothing to do with CAMB. This kink shows up at all redshifts I've tried (but it moves around). Notice that it is present in the linear spectrum as well. It agrees much better after the kink than before, does CAMB change methods at some l? The kink might be from splining between methods.

This was calculated with all three accuracy settings set to 3. Decreasing these brings the kink to smaller l. Is there an something obvious I'm doing wrong here?

Code: Select all

output_root = results/
get_scalar_cls = T
get_vector_cls = F
get_tensor_cls = F
want_CMB = F
num_redshiftwindows = 1
DoRedshiftLensing = F
limber_windows = T
Do21cm = F
redshift(1) = 1.75
redshift_kind(1) = lensing
redshift_sigma(1) = 0.75
COBE_normalize = F
CMB_outputscale = 1
get_transfer = T
do_nonlinear = 2
l_max_scalar = 13000
k_eta_max_scalar = 50000
do_lensing = T
lensing_method = 1
w = -1
cs2_lam = 1
hubble = 0.719E+02
use_physical = T
ombh2 = 0.2280E-01
omch2 = 0.1106E+00
omnuh2 = 0
omk = 0
temp_cmb = 2.726
helium_fraction = 0.24
massless_neutrinos = 3.04
massive_neutrinos = 0
nu_mass_eigenstates = 1
nu_mass_degeneracies = 1
nu_mass_fractions = 1
transfer_high_precision = T
reionization = T
re_use_optical_depth = T
re_optical_depth = 0.87E-01
RECFAST_fudge = 1.14
recombination = 1
initial_power_num = 1
scalar_spectral_index(1) = 0.963E+00
scalar_nrun(1) = 0
scalar_amp(1) = 2.2387e-009
initial_condition = 1
scalar_output_file = scalCls.dat
accurate_polarization = F
accurate_reionization = F
accurate_BB = T
do_late_rad_trunction = T
do_late_rad_trunction = T
do_tensor_neutrinos = F
feedback_level = 1
massive_nu_approx = 3
number_of_threads = 0
accuracy_boost = 3
l_accuracy_boost = 3
l_sample_boost = 3

[Antony Lewis]

Most likely to be the limber_windows setting.

There are a few bugs in the current CAMB sources too. I can give you a link for the current test build if problems persist.

[kiyo masui]

That did it. Many thanks.