Early ionization history: RECFAST2

Post Reply
Antony Lewis
Posts: 1941
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Early ionization history: RECFAST2

Post by Antony Lewis » September 26 2004

RECFAST (astro-ph/9909275) is used by CAMB and CMBFAST to compute the background ionization history (without reionization). It needs to be accurate for future CMB observations - RECFAST changes the C_l relative to Peebles recombination by a few percent on small scales.

Has anyone independently checked out this paper and code? I have checked almost all of CMBFAST (i.e. re-wrote in CAMB), but RECFAST is currently shared and unchecked. That it is correct and accurate will be important!

CMBFAST has been checked against a full Boltzmann code for simple models (astro-ph/0306052), however this accuracy test was contingent on RECFAST being correct.

Any work done or planned on this? (checking new rates data, testing accuracy and fudge parameters, etc)

Anze Slosar
Posts: 183
Joined: September 24 2004
Affiliation: Brookhaven National Laboratory
Contact:

recfast

Post by Anze Slosar » September 27 2004

Incidentally, I just found this comment in the rec-fast code:

!CD The Cosmological model can be flat or open.

Does this pose any problems for closed models? Anyway, shouldn't matter too much for nearly-closed models...

While you are answering this, you can also answer what I was really looking for: how do I get mean redshift of recombination from cosmomc/camb? :)

Antony Lewis
Posts: 1941
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

recfast

Post by Antony Lewis » September 29 2004

In CAMB at least the RECFAST code has been modified to get the background H evolution from the global [tex]d\tau/da[/tex] function, which will handle any geometry correctly.

In any case the effect of non-zero curvature should presumably be pretty much negligible before recombination.

There is a variable tau_maxvis defined in modules.f90 which is the (approximate I think) maximum of the visibility in conformal time. [this question should be in the software forum!]

Michael Doran
Posts: 41
Joined: November 22 2004
Affiliation: ITP Heidelberg
Contact:

Early ionization history: RECFAST2

Post by Michael Doran » November 25 2004

Hi,

first of all, I re-wrote RECFAST replacing all
hard-wired background quantities by the ones computed in the central COSMOS class of CMBEASY. So that's more or less what you did, Antony, I suppose.

Even better: In the summer 2003, I wrote (and I ship it with cmbeasy) a version of Recfast which can be used for varying fine structure constants. As the original RECFAST is hard-wired in many places (i.e. when does some process stop etc.), I had to re-write RECFAST completly and on my way think very thoroughly through all the steps. So this class of cmbeasy (I believe I called it RecfastAlpha) is an independent implementation of the equation system (I even looked at the reaction rate papers the original authors used [if I recall correctly, the extrapolation from that laboratory data is a bit frigthening, but my memory is hazy]).

For on-line documantation on that class, see
http://www.thphys.uni-heidelberg.de/~do ... Alpha.html

So yes, there is an independent implementation of RECFAST and as far as I recall, they yield similar results (just by setting delta alpha=0). However, I never checked extensivly.

BTW: CMBEASY uses a completly independent "time stepping" through recombination and re-ionization compared to CMBFAST (and most likely CAMB, though I'm no expert). Again, the results are quite close. However, one may always trade accuracy for speed by choosing smaller step sizes for the perturbation evolution of each mode.

Michael

Antony Lewis
Posts: 1941
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: Early ionization history: RECFAST2

Post by Antony Lewis » November 26 2004

This is good. Would be useful to know how accurately you agree. (I believe the CMBFAST folk have also checked the code relative to the paper.)

However I'd really like to see an independent check of the RECFAST paper, updating of rates from new papers, and parameterization of any 'frightening' extrapolations or remaining uncertainties so we can see what affect uncertainties have on the results, and if neccessary allow them to be used as parameters to be marginalized over.

Michael Doran
Posts: 41
Joined: November 22 2004
Affiliation: ITP Heidelberg
Contact:

Early ionization history: RECFAST2

Post by Michael Doran » November 26 2004

Hi,
I believe the CMBFAST folk have also checked the code relative to the paper
I don't get this. In what sense ? Can you explain in more detail ?

I'm no profound expert on recombination and thus out of the top of my head would not be able to improve on the equations that recfast uses. However, estimating the uncertainties from the data they use might be much easier.

We could also compare recfast to recfastalpha, then we would sort of know the implementation uncertainty...

Any suggestions ?

Michael

Antony Lewis
Posts: 1941
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: Early ionization history: RECFAST2

Post by Antony Lewis » November 26 2004

I think someone checked the RECFAST code to make sure what was implemented in the code was the same as was written in the paper. There were some fairly minor corrections (also in CAMB) a while ago. But they didn't check the paper.

Perhaps you could output a reionization history from CMBFAST/CAMB and compare with what you get?

Post Reply