Arbitrary source planes

[damonge]

Allows for lensing (and ISW) source planes to lie outside of the box. Line-of-sight integrals will only contain information up to z=z_max (passed in the param file), but the radial kernel will be that of a source plane at the requested distance.

Closes #75

[andreufont]

I don't know enough about the code to validate the changes, but I'm sure you got it right.

@Laura-Casas - could you try running a new box with this branch, before merging?

We can then ask @dljarque to validate the new box.

[dljarque]

Here is the validation for CMB lensing of two CoLoRe boxes run by @Laura-Casas. One prior to the source plane z_s change, and the second one with the new implementation of this pull request. They both seem ok, thanks!

Theory lines are computed with the linear PK from CAMB (matching CoLoRe cosmology) and manually integrating the Limber approximation up to z_cut.

[damonge]

Thanks @dljarque ! Any reason why you had to set zs=1090 in one case and 1100 in the other?

[dljarque]

@damonge CoLoRe boxes were run with z_out = 1100.0. However, when setting the same cosmology in CAMB (for the theory lines) as the CoLoRe boxes, the z_star parameter that outputs is around 1090. Should they be the same?

[andreufont]

Well, it would be nice if for the next boxes, we set z_out in the config file to whatever is the expected value of z_star for the input cosmology, so that we are consistent.

[dljarque]

I am adding an updated version of the previous plot. I am validating just the second CoLoRe box now but adding a smoothing parameter $k_p$ into the theory linear PK, so that is multiplied by a damping exponential given by the square of $k/k_p$. There is also the box relative error wrt the theory line from CAMB, split into two ranges, $L<100$ and $L \geq 100$, to better see the deviation at different scales.

Furthermore I am also adding a plot of the cumulative (on multipole L) chi2 estimator as a function of the multipole L, along with the cumulative p-value of the fit. You can clearly see how it diverges from a good fit at very high L, when the smoothing parameter no longer matches the theory. The cosmic variance noise of the lensing field is really low there which makes chi2 even bigger.