Skip to content
This repository has been archived by the owner on Apr 5, 2023. It is now read-only.

Tensor modes compatible with the Websky CMB #48

Closed
zonca opened this issue Feb 12, 2020 · 31 comments
Closed

Tensor modes compatible with the Websky CMB #48

zonca opened this issue Feb 12, 2020 · 31 comments
Labels
good first issue Good for newcomers help wanted Extra attention is needed
Projects
[pwg-mbs]: Map-based simulations

Comments

@zonca
Copy link
Member

zonca commented Feb 12, 2020

@marcelo-alvarez @ajvanengelen Would it be possible for you to run a simulation of just the primordial B modes (no lensing) with the same cosmological parameters used for the WebSky CMB simulations?
Using a fiducial value of r (any would work, e.g. 1e-3).

Then I can add an option to the WebSkyCMB to specify a value of r and add those primordial B modes properly weighted.
@ajvanengelen
Copy link
Contributor

ajvanengelen commented Feb 12, 2020 via email •
edited

@zonca
Copy link
Member Author

zonca commented Feb 13, 2020

@ajvanengelen @keskitalo I think we can keep the TT low-ell component as well.

Yes, thanks, just C_{ell} is enough

@zonca
Copy link
Member Author

zonca commented Feb 25, 2020

@ajvanengelen any news about this?

@zonca
Copy link
Member Author

zonca commented Apr 1, 2020

@ajvanengelen would you have any update about this?

@ajvanengelen
Copy link
Contributor

ajvanengelen commented Apr 1, 2020 via email

@zonca
Copy link
Member Author

zonca commented Apr 1, 2020

no

@ajvanengelen
Copy link
Contributor

ajvanengelen commented Apr 1, 2020 via email

@zonca
Copy link
Member Author

zonca commented Apr 1, 2020

From @ajvanengelen:

NERSC:

/global/cscratch1/sd/engelen/websky_BB_from_tensors_r_eq_1.npz.npy

This is the BB spectrum for the Websky cosmology from a model with r = 1 (which of course needs to be scaled to whatever actual r value we want to use). It is in the format that pixell expects for power spectra, namely an array of shape (3, 3, N_ell) and so the spectrum is in the [2, 2, :] elements. Units are plain C_l in uK^2.

@zonca
Copy link
Member Author

zonca commented Apr 1, 2020

thanks @ajvanengelen!
I'll implement this into the model

@zonca zonca added this to In progress in [pwg-mbs]: Map-based simulations Apr 6, 2020
@zonca zonca added good first issue Good for newcomers help wanted Extra attention is needed labels Apr 7, 2020
@zonca
Copy link
Member Author

zonca commented Apr 7, 2020

if anyone is willing to help implement this:

@ajvanengelen
Copy link
Contributor

ajvanengelen commented Apr 7, 2020 via email

@ajvanengelen
Copy link
Contributor

ajvanengelen commented Apr 7, 2020 •
edited

Just a note that this is a bit of a heterogeneous mixture of components. In the current plan, these B modes from tensors will not be lensed, unlike the E and T modes. Furthermore, we are not making any T fluctuations from these tensors.

@zonca
Copy link
Member Author

zonca commented Apr 23, 2020 •
edited

@ajvanengelen I am implementing this, so here is a plot of your spectra:

image

So just to confirm, to add tensor modes to the existing lensed/unlensed CMB spectra which are already in so_pysm_models (https://so-pysm-models.readthedocs.io/en/latest/models.html#websky) I just use the BB spectrum (and scale properly by r) from this dataset and discard the rest. Right?

@ajvanengelen
Copy link
Contributor

ajvanengelen commented Apr 24, 2020 via email

@zonca
Copy link
Member Author

zonca commented May 14, 2020

@ajvanengelen I have implemented this and ran some simulations, however, as it is evident from the plot above, it looks like there is no more power after ell=600, even if your files are until ell=8000 or so.

See an example simulation looks like this:

image

do you mind to produce another file with power all the way to 8000?

@ajvanengelen
Copy link
Contributor

ajvanengelen commented May 16, 2020 •
edited

I am not sure why this is happening, as I do pars.set_for_lmax(10000, lens_potential_accuracy=2) in my CAMB run. I don't see a separate function to set the accuracy parameters differently for tensors vs. scalars. Given how strongly the primordial BB signal is suppressed on scales smaller than the horizon scale at decoupling, though, does this lmax=600 matter in practice?

@zonca
Copy link
Member Author

zonca commented May 16, 2020

I don't know, you are the expert, you confirm lmax=600 is fine, I will write that in the docs and be done.

@ajvanengelen
Copy link
Contributor

ok :) let's write that in the docs, hopefully it is ok. Sorry for the confusion about this!

@zonca
Copy link
Member Author

zonca commented May 16, 2020

ok, thank you!

@zonca zonca closed this as completed May 19, 2020
[pwg-mbs]: Map-based simulations automation moved this from In progress to Done May 19, 2020
@zonca zonca reopened this May 21, 2020
[pwg-mbs]: Map-based simulations automation moved this from Done to In progress May 21, 2020
@keskitalo
Copy link
Member

keskitalo commented May 21, 2020 •
edited

@mtristram just discovered that scaling the tensor modes to match a specific value of r is inaccurate because the tensor spectral index, n_t, is set by the consistency relation: n_t = -r/8. When the tensor mode spectrum is scaled, the spectral index is not adjusted to match the new r. This is particularly problematic if the fiducial value of r is chosen as 1 and not some small, more representative value.

@ajvanengelen
Copy link
Contributor

Whoops - I agree that this is an issue with these spectra. I didn't realize that if n_t is unspecified in the python camb, it is set to be -r/8. I will do a new run with n_t set to 0.

Given that I will be rerunning this, Andrea, what is your preferred data format? I sent numpy.save files last time, but would you prefer .fits?

@zonca
Copy link
Member Author

zonca commented May 21, 2020

best is the same format as last time, I have already a notebook that does the conversion

@zonca
Copy link
Member Author

zonca commented May 28, 2020

@ajvanengelen any update on this?

@ajvanengelen
Copy link
Contributor

Sorry for the delay, please have a look now at /global/cscratch1/sd/engelen/websky_BB_from_tensors_r_eq_1.npz.npy on nersc/cori.

@zonca zonca mentioned this issue Jun 2, 2020
Closed
11 tasks
@zonca
Copy link
Member Author

zonca commented Jun 2, 2020

thanks @ajvanengelen see comparison between new and old spectra

image

also changed the scaling to compare more easily with something like: https://lambda.gsfc.nasa.gov/graphics/bb_upperlimits/bb_limits_2015apr.pdf

zonca added a commit that referenced this issue Jun 2, 2020
@zonca
mention n_t
4bb9d67 
see
#48 (comment)
@zonca
Copy link
Member Author

zonca commented Jun 2, 2020

ok, I updated the cl fits file at https://portal.nersc.gov/project/cmb/so_pysm_models_data/websky/0.3/

@zonca zonca closed this as completed Jun 2, 2020
[pwg-mbs]: Map-based simulations automation moved this from In progress to Done Jun 2, 2020
@zonca
Copy link
Member Author

zonca commented Jun 18, 2020

@ajvanengelen can you please attach here or put in a gist the script you used to generate this?

@zonca zonca reopened this Jun 18, 2020
[pwg-mbs]: Map-based simulations automation moved this from Done to In progress Jun 18, 2020
@ajvanengelen
Copy link
Contributor

Sure - https://github.com/ajvanengelen/webskylensing/blob/master/py/test_cmb_powerspectra.py - as you can see, this calls my get_cmb_powerspectra.websky_cmb_spectra routine which is in the same directory.

It is just a wrapper around CAMB that sets the cosmo parameters to be those used in the websky sims, and then (I think) packages the results into the ordering that the pixell tools expect for making alm's.

@zonca
Copy link
Member Author

zonca commented Jun 29, 2020

ok thanks

@zonca zonca closed this as completed Jun 29, 2020
[pwg-mbs]: Map-based simulations automation moved this from In progress to Done Jun 29, 2020
@zonca zonca reopened this Dec 10, 2020
[pwg-mbs]: Map-based simulations automation moved this from Done to In progress Dec 10, 2020
@zonca
Copy link
Member Author

zonca commented Dec 10, 2020

@ajvanengelen I am running some checks on the CMB tensor modes,
I wanted to check with you my implementation:

you simulated the CMB tensor modes with r = 1, if I want to simulate instead r=1e-3, I should multiply the TT, EE, BB power spectrum (not the map) by 1e-3, correct?

@zonca
Copy link
Member Author

zonca commented Jan 8, 2021

correct, r is defined as ratio of the C_ell so it is in power, if I am weighting the map, I use the sqrt.

@zonca zonca closed this as completed Jan 8, 2021
[pwg-mbs]: Map-based simulations automation moved this from In progress to Done Jan 8, 2021
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Assignees
No one assigned
Labels
good first issue Good for newcomers help wanted Extra attention is needed
Projects
No open projects
[pwg-mbs]: Map-based simulations
  
Done
Milestone
No milestone
Development

No branches or pull requests
3 participants
@zonca @keskitalo @ajvanengelen