TCA being unstable when both l_max_ur and l_max_ncdm are increased

[nanoomlee]

Hi all,

I have recently noticed that some numerical errors arise during tight-coupling approximation when both l_max_ur and l_max_ncdm are increased from their default value. This issue does not appear when only one of them is increased and the other is kept at its default value.

In the attached plot, the differences between theta_b and theta_g are shown extracted from perturbations_tca_slip_and_shear() function, for two configurations of (l_max_ur, l_max_ncdm) = (17, 17) and (30, 30). For this test, I used the most recent version, and used explanatory.ini setting N_eff=2.044 and N_ncdm=1.

While the resulting error in matter power spectrum at late times is small, the behavior itself is clearly unphysical and I would like to address it. Since this issue also appears to be related to what I am currently working on in my modified version of CLASS, finding a fix would be very helpful.

Could you please let me know if there is a straightforward way to resolve this? Thank you very much.

TCA_theta_b_theta_g.pdf