CMBFAST Trouble with nonflat universes

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply
Matthew Kerr
Posts: 3
Joined: April 23 2007
Affiliation: University of Washington

CMBFAST Trouble with nonflat universes

Post by Matthew Kerr » April 24 2007

Hi all,

I am having some difficulties getting CMBFast to spit out power spectra for non-flat geometries. I am doing very simple cosmologies (w= -1, scalar perturbations only, no lensing, pretty near standard LCDM fiducial model, adiabatic initial conditions), but my power spectra all come back NaN. Flat cosmologies are working fine. To add to the puzzle, when I use the same parameter values and conditions on NASA's web-based CMBFast implementation, everything works, so I'm convinced there is some problem with my spherical Bessel function generator/file.

Has anyone else experienced the NaN phenomenon in the output? Cheers,

Matthew

Matthew Kerr
Posts: 3
Joined: April 23 2007
Affiliation: University of Washington

CMBFAST Trouble with nonflat universes

Post by Matthew Kerr » April 25 2007

Following up after some more experimentation... I'm positive something's wrong with my use of the ultraspherical Bessels, but I can't figure it out. Here's a recent open universe I tried:

0
5 200
1
-1
0.05 0.2 0.7 0.0
72.5 2.726 0.24 3.04 0 10.75
0
0
0
1 1 0
test_curved
1
ujl_500_500

I've played around with plenty of lmax/kmax pairs both in the CMB parameters and in the ultraspherical generator. I almost believe there's a bug in the ujlgen, but that seems hard to believe. Any thoughts?

Ben Gold
Posts: 81
Joined: September 25 2004
Affiliation: University of Minnesota
Contact:

CMBFAST Trouble with nonflat universes

Post by Ben Gold » April 26 2007

I tried your parameters and got the following output

Code: Select all

    2  0.57106E-01  0.79187E-10  0.11164E-05
    3  0.56150E-01  0.28635E-09  0.27074E-05
    4  0.54529E-01  0.76124E-09  0.47678E-05
    5  0.53201E-01  0.16779E-08  0.72542E-05
You need to run ujlgen with the exact same lmax,kmax that you intend to give cmbfast, and use that specific ujl file. I don't think you can run ujlgen with a high lmax,kmax and use smaller values in cmbfast without confusing the code (generally it just crashes). Otherwise I really don't know what you could be doing differently.

Matthew Kerr
Posts: 3
Joined: April 23 2007
Affiliation: University of Washington

CMBFAST Trouble with nonflat universes

Post by Matthew Kerr » April 26 2007

Thanks for checking that out for me. I just double-checked with an ultraspherical Bessel generated as you suggested, and I'm still getting NaN. Guess I'm going to have to see why Fortran libraries are broken on our network! Thanks again.

Akhilesh Nautiyal(akhi)
Posts: 72
Joined: June 13 2007
Affiliation: Malaviya National Institute of Technology Jaipur

CMBFAST Trouble with nonflat universes

Post by Akhilesh Nautiyal(akhi) » June 18 2007

Hi,

First of all again untar the file cmbfast.tar.
Go to cmbfast.inc and change the value of nstep0 to
parameter (nstep0=8000).
Coppy the file lvalues_fine.inc to lvalues.inc.
Then configure cmbfast again and run make.
After that follow the process of runing jlgen, ujlgen etc as for flat case.
Then run cmbfast with curved geometry.
I hope it will work.

Regards,
Akhilesh

Ivan Debono
Posts: 25
Joined: April 17 2007
Affiliation: APC, Paris 7 / OBSPM

CMBFAST Trouble with nonflat universes

Post by Ivan Debono » June 19 2007

I've got a problem which could be related. I'm trying to run CMBfast with the following universe, but apparently you can't run CMBfast with a pure CDM universe:

2
1500 3000
5 5
1 0
cmb.tf
1
-0.99
0 1 0 0 #i.e. omega_CDM=1; all the other omegas=0
50 2.726 0.24 2 1 10.75
1
1
0.166
0
1 0.99 0.0
0
cmb.cl
1
jlgen.dat

Has anyone else encountered the same problem? Am I doing something wrong?

Ben Gold
Posts: 81
Joined: September 25 2004
Affiliation: University of Minnesota
Contact:

CMBFAST Trouble with nonflat universes

Post by Ben Gold » June 19 2007

I suspect that the main problem is that you're asking it to do dark energy perturbations with [tex]w=-0.99[/tex] even though there's no dark energy present. At the very least try turning perturbations off. I don't think CMBFAST is smart enough to detect cases where the d.e. perturbations become pathological.

Another possibility is that you're asking for a massive neutrino even though you've set [tex]\Omega_\nu[/tex] to zero, but I think that particular error isn't fatal to CMBFAST. But if it still doesn't work you could try a small nonzero [tex]\Omega_\nu[/tex] of [tex]10^{-3}[/tex] or so.

Ivan Debono
Posts: 25
Joined: April 17 2007
Affiliation: APC, Paris 7 / OBSPM

CMBFAST Trouble with nonflat universes

Post by Ivan Debono » June 19 2007

Thanks. I'll try playing about with some of these parameters. I suspect the D.E. parameter to be the cause of the problem.

Ivan Debono
Posts: 25
Joined: April 17 2007
Affiliation: APC, Paris 7 / OBSPM

CMBFAST Trouble with nonflat universes

Post by Ivan Debono » June 26 2007

Still haven't managed to solve it. Is there any way in which one could use cmbfast with a pure-CDM cosmological model?

Post Reply