[yt-users] r_min in HaloProfiler
Matthew Turk
matthewturk at gmail.com
Tue Dec 8 21:10:25 PST 2009
Hi everyone,
I think we've gotten fairly far off topic at this point, and further
discussions should be carried out on the enzo users list or off-list.
-Matt
On Tue, Dec 8, 2009 at 9:07 PM, Michele Trenti
<Michele.Trenti at colorado.edu> wrote:
> Hi Shankar,
>
>> Actually, I don't care for the resolution of the halos. I am analyzing the
>> suppression of matter power spectrum as a function of neutrino mass. Which
>> means I am interested only on the semi-linear scales of 20-60 Mpc/h. Towards
>> this, I wish to extract small galaxy groups (upto 1e13M_solar) from my
>> simulation. So, can I regard the HaloFinder halos as blobs, w/o caring for
>> the resolution ? If yes, can I include halofinder halos with as low as
>> 50-100 particles (since they are still halos, regardless of the resolution)
>> ?
>>
>> I will appreciate if I can get some opinion on this.
>
> I would strongly reccomend to read O'Shea et al. 2005, ApJS, 160, 1. The
> halo mass function in Enzo is not correctly reproduced below ~200-300
> particles, even if you use AMR. You start missing a significant number of
> halos below that threshold. This is because the use of a PM scheme at very
> high z (at the beginning of your simulation) suppresses high frequency modes
> in the power spectrum. If you need to study the halo mass function/power
> spectrum, then I would suggest you to consider an alternative code. For
> example, Gadget2 (Springel 2005) can reproduce correctly the halo mass
> function down to ~30 particles and might be better suited to your needs.
> Like Enzo, it is public.
>
> Cheers,
>
> Michele
>
> Michele Trenti, Ph.D.
> Center for Astrophysics and Space Astronomy
> University of Colorado
> 389 UCB Phone: +1 303-492-8490
> Boulder, Colorado 80309-0389 Fax: +1 303-492-7178
> USA E-mail: michele.trenti at colorado.edu
>
>
> "It often happens that a player carries out a deep and complicated
> calculation, but fails to spot something elementary right at the first
> move." -- Chess Grandmaster Alexander Kotov
>
>
>
> On Tue, 8 Dec 2009, Agarwal, Shankar wrote:
>
>> Eric and Sam,
>>
>> Yeah. that makes sense. It boils down to the resolution of the box.
>>
>> A question to Sam ... How did you get "mass resolution is about
>> 2e10M_solar" ?
>>
>> I am getting 5e9M_solar as the mass of each dark matter particle. And the
>> number of particles in a 1e13M_solar halo is 2000.
>>
>> ID mass particles density
>> x y
>> z centreofmass x centerofmass y
>> centerofmass z
>> 3700 9.853250284e+12 2000 1.136880784e+04 3.619545698e-01
>> 4.593685567e-01 8.036487699e-01 3.618929449e-01 4.593253520e-01
>> 8.041266519e-01
>>
>>
>> So I guess for 250 particles, I can try...
>>
>> virial_filters=[['TotalMassMsun','>=','1.25e12']]
>>
>>
>>
>> A question to Eric :
>>
>> Actually, I don't care for the resolution of the halos. I am analyzing the
>> suppression of matter power spectrum as a function of neutrino mass. Which
>> means I am interested only on the semi-linear scales of 20-60 Mpc/h. Towards
>> this, I wish to extract small galaxy groups (upto 1e13M_solar) from my
>> simulation. So, can I regard the HaloFinder halos as blobs, w/o caring for
>> the resolution ? If yes, can I include halofinder halos with as low as
>> 50-100 particles (since they are still halos, regardless of the resolution)
>> ?
>>
>> I will appreciate if I can get some opinion on this.
>>
>> shankar
>>
>>
>>
>>
>> -----Original Message-----
>> From: yt-users-bounces at lists.spacepope.org on behalf of Eric Hallman
>> Sent: Tue 12/8/2009 4:53 PM
>> To: Discussion of the yt analysis package
>> Subject: Re: [yt-users] r_min in HaloProfiler
>>
>> Shankar,
>> sam is right on the money with his analysis. Although the halo
>> finders identify a large number of halos in your simulation they are
>> very poorly resolved, and so this filter will rule them out on the
>> basis of not getting to OD of 200. Our analysis with enzo shows that
>> you can not expect the halos to be resolved under a few hundred
>> particles as sam suggests.
>>
>> I would take a look at how many halos are above 500 particles or so
>> and only run your analysis on those. I suspect it is possible that
>> one of the reasons you are not seeing speedup with the parallel runs
>> is that you don't have enough work for the processors given the small
>> number of virialized halos that you'll actually analyze with this
>> filter.
>>
>>
>> On Dec 8, 2009, at 3:30 PM, Sam Skillman wrote:
>>
>>> Shankar,
>>>
>>> In a box size of 200 Mpc/h^3 and 512^3 particles, your mass
>>> resolution is about 2e10M_solar. That means in a 10^13 M_solar
>>> halo, you have at most ~500 particles in your halos, which is well
>>> resolved. The main problem with what you're doing is putting an
>>> upper limit on the mass of the halos that you are profiling, which
>>> is opposite to what one normally does. Even if you want to be
>>> cavalier with what you call a halo, you definitely don't want to go
>>> much below 2-5e12 since then you're talking 100 particles. I
>>> personally don't trust profiles unless they have a few thousand
>>> particles, because then I might believe the hydro is resolved. My
>>> guess is that if you do:
>>>
>>> virial_filters=[['TotalMassMsun','>=','5e12']]
>>>
>>> you'll have more luck with them actually being virialized.
>>>
>>> Main point:
>>> If you want to study halos with masses less than 10^13, you're using
>>> the wrong simulation (i.e. too large of a physical box size or too
>>> small of a grid).
>>>
>>> Sam
>>>
>>>
>>> On Tue, Dec 8, 2009 at 3:04 PM, Agarwal, Shankar <sagarwal at ku.edu>
>>> wrote:
>>> Sam,
>>>
>>> I am trying to get the virial masses of the halos (with mass upto 1e
>>> +13 Msun) with this filter...
>>>
>>> hp.add_halo_filter(HP.VirialFilter, must_be_virialized=True,
>>> overdensity_field='ActualOverdensity',
>>> virial_overdensity=200,
>>> virial_filters=[['TotalMassMsun','<=','1e13']],
>>> virial_quantities=['TotalMassMsun','RadiusMpc'])
>>>
>>>
>>> But when I look at the radial_profiles/Halo_xxxx_profile.dat files,
>>> I see that none of the halos are virialized. The peak halo densities
>>> are reaching about 20*mean_matter_density. I began my 200Mpc/h Box
>>> with 512^3 particles simulation at z=99. I have not looked at the
>>> literature in great detail but is there something fishy here ?
>>>
>>>
>>> shankar
>>>
>>>
>>>
>>> -----Original Message-----
>>> From: yt-users-bounces at lists.spacepope.org on behalf of Sam Skillman
>>> Sent: Tue 12/8/2009 3:23 PM
>>> To: Discussion of the yt analysis package
>>> Subject: Re: [yt-users] r_min in HaloProfiler
>>>
>>> Shankar,
>>>
>>> r_min is defined in your email:
>>> r_min = 2 * self.pf.h.get_smallest_dx() * self.pf['mpc']
>>>
>>> it's twice the smallest dx in units of mpc, dx being the smallest
>>> cell size.
>>> it is being rejected because your halo has something like 3 cells
>>> (radial)
>>> in it, which is almost certainly not resolved.
>>>
>>> sam
>>>
>>> On Tue, Dec 8, 2009 at 2:11 PM, Agarwal, Shankar <sagarwal at ku.edu>
>>> wrote:
>>>
>>>> Hi,
>>>>
>>>> While running HaloProfiler, I am seeing this...
>>>>
>>>> P001 yt.lagos ERROR 2009-12-08 15:53:22,403 Skipping halo
>>>
>>> with r_max
>>>>
>>>> / r_min = 1.885682.
>>>>
>>>>
>>>>
>>>> I looked at yt/extensions/HaloProfiler.py...
>>>>
>>>> r_min = 2 * self.pf.h.get_smallest_dx() * self.pf['mpc']
>>>> if (halo['r_max'] / r_min < PROFILE_RADIUS_THRESHOLD):
>>>> mylog.error("Skipping halo with r_max / r_min =
>>>
>>> %f." %
>>>>
>>>> (halo['r_max']/r_min))
>>>>
>>>>
>>>> r_max would be the distance to the furthest particle in the halo
>>>
>>> as found
>>>>
>>>> by HaloFinder.
>>>> But what is r_min ? And what is the basis for skipping this halo ?
>>>>
>>>>
>>>> shankar
>>>> _______________________________________________
>>>> yt-users mailing list
>>>> yt-users at lists.spacepope.org
>>>> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>>>>
>>>
>>>
>>>
>>> --
>>> Samuel W. Skillman
>>> DOE Computational Science Graduate Fellow
>>> Center for Astrophysics and Space Astronomy
>>> University of Colorado at Boulder
>>> samuel.skillman[at]colorado.edu
>>>
>>>
>>> _______________________________________________
>>> yt-users mailing list
>>> yt-users at lists.spacepope.org
>>> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>>>
>>>
>>>
>>>
>>> --
>>> Samuel W. Skillman
>>> DOE Computational Science Graduate Fellow
>>> Center for Astrophysics and Space Astronomy
>>> University of Colorado at Boulder
>>> samuel.skillman[at]colorado.edu
>>> _______________________________________________
>>> yt-users mailing list
>>> yt-users at lists.spacepope.org
>>> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>>
>> Eric Hallman
>> Google Voice: (312) 725-HMAN
>> hallman13 at gmail.com
>>
>>
>>
>>
>>
> _______________________________________________
> yt-users mailing list
> yt-users at lists.spacepope.org
> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>
More information about the yt-users
mailing list