[yt-users] potential unit error in cosmology.py

Britton Smith brittonsmith at gmail.com
Tue Aug 9 04:40:09 PDT 2016


Hi Pengfei,

Firstly, Nathan, JohnZ, Matt, Cameron, if you could have a look at the last
paragraph of this response, I would appreciate it.

Ah, I see now.  I actually get a different answer for the redshift interval
when I do it by hand.  I get 9.994 instead of 9.979.  Still, our answers,
differ from the value calculated by the code for the reason you describe,
essentially that the comoving to proper conversion is different between a
cosmology object created from a dataset (where a specific redshift is
defined) and one created by hand.  In the one created from a dataset, the
comoving and proper frame differ by the factor of (1+z), whereas in the one
created by hand, they are the same for the reasons I described in my
previous email.

This is definitely a problem.  We cannot be doing cosmology distance
calculations in two different unit systems, one where comoving and proper
are the same and one where they are different.  As strange as this may
sound, the one that is least correct is using the system where the comoving
and proper frames are different.  That's because something like a comoving
radial distance can only be expressed in the comoving frame and has no
direct translation to the proper frame via some factor of (1 + z).

The solution to this, as I see it, is the following:
The cosmology object that gets associated with a dataset must have its
unit_registry overridden to set the comoving and proper frames equal to
each other.  A better solution would be to even remove the proper frame
entirely from the cosmology object's unit_registry, though I don't even
know if that's possible.  I would appreciate the input of someone with more
intimate knowledge of the unit system, like Nathan, Matt, or John Z.  As a
quick test, I experimented with the case where we create a cosmology object
from a dataset by passing a copy of the unit_registry and then modifying
the comoving frame.  I ran into some difficulties because it looks like,
for example, Mpc is define independently of pc, and so I would have to
modify both pccm and Mpcm, and perhaps a number of others that I am unaware
of.  If anyone has ideas on how best to do this, please say so.

In conclusion, I do think there is an issue that needs to be fixed.  Input
here would be appreciated.  Thanks, Pengfei, for identifying this issue!

Britton

On Mon, Aug 8, 2016 at 10:25 PM, Pengfei Chen <madcpf at gmail.com> wrote:

> Hi Britton,
>
> Thanks very much for your test! The error happens when I make light rays
> from one single dataset. I tried to reproduce the error with the dataset
> under src/yt-hg/tests. The code I used is here:
>
> http://paste.yt-project.org/show/6751/
> After running the code, I did
>
> h5ls -d lightray.h5/grid/redshift
> and found the last redshift was 9.98999107782409. However, the correct
> value should be ~ 9.979.
>
> When I call LightRay this way, it will call the _deltaz_forward function
> in cosmology_splice.py, which will call the comoving_radial_distance in
> cosmology.py:
>
>         distance2 = self.cosmology.comoving_radial_distance(z2, z)
> , where distance2 is not in comoving units. Then it will calculate z2:
>
>             z2 = ((target_distance - distance2) / m.in_units("Mpccm / h"))
> + z2
>
> . I think the subtraction here is the problem, since it tries to convert
> distance2 to comoving units, even though the numerical value of distance2
> is already comoving. So a wrong factor of (1+z) is multiplied.
>
> Any comment is appreciated.
>
> Thanks again for your help!
> Pengfei
>
> On Mon, Aug 8, 2016 at 7:10 AM, Britton Smith <brittonsmith at gmail.com>
> wrote:
>
>> Hi Pengfei,
>>
>> I'm not able to reproduce what you're seeing, but before I go into that,
>> I can explain a few things about how the cosmology calculator works.  In
>> yt, we are only able to define a comoving unit system in relation to a
>> proper unit system at a specific redshift.  We are unable to define a
>> comoving unit system that is the base unit system.  This is a problem for
>> the cosmology calculator, which only calculates comoving quantities.
>> Therefore, in this case, we are forced to set the comoving and proper unit
>> systems to be the same.  You can see this for yourself by running the
>> following script:
>>
>> from yt.utilities.cosmology import Cosmology
>> co = Cosmology()
>> print (co.comoving_radial_distance(1.92, 2).to("Mpc/h"))
>> print (co.comoving_radial_distance(1.92, 2).to("Mpccm/h"))
>>
>> Both will be the same answer.
>>
>> Now, back to the issue you're seeing.  I ran the following script:
>> http://paste.yt-project.org/show/6741/
>> to calculate the redshifts dumps needed for a 80 Mpc/h box to go from z =
>> 2 to z = 1.92 and got the following result:
>> CosmologyOutputRedshift[0] = 2.000
>> CosmologyOutputRedshift[1] = 1.926
>>
>> This is pretty close to what you expected.  Additionally, if I run the
>> following script,
>> http://paste.yt-project.org/show/6743/
>> with the above redshifts added to the parameter file, I get that dz_max
>> (the change in redshift when moving 80 Mpccm/h at z = 0)
>> of 0.0746521607364, which also seems to agree with what you are finding.
>>
>> Can you share with me the simulation parameter file and script that
>> you're using to make your light rays?  I can take a look at that and see if
>> I still get what I expect.
>>
>> Britton
>>
>> On Mon, Aug 8, 2016 at 6:13 AM, Pengfei Chen <madcpf at gmail.com> wrote:
>>
>>> Hi all,
>>>
>>> I was trying to generate light rays using yt 3.4-dev, but found that the
>>> light rays have wrong redshift intervals. For example, my simulation box is
>>> 80Mpccm/h per side, and I expect a light ray generated in a data dump at
>>> z=2 with length =1 simulation unit (i.e. 80Mpccm/h) would have a redshift
>>> interval from z=2 to z=1.92. However, the light ray generated by yt gives
>>> spans from z=2 to z=1.97.
>>>
>>> As I look into this, I found that the function comoving_radial_distance
>>> in cosmology.py might return the wrong unit. I think the return value
>>> should be in comoving units, instead of physical units. To see this
>>> directly, I made the following change in cosmology_splice.py:
>>>
>>> (root) ~/yt-conda/src/yt-hg/yt/analysis_modules/cosmological_observation
>>> $diff cosmology_splice.py cosmology_splice.py0
>>> 373d372
>>> <             print target_distance, distance2, z2
>>>
>>> And it shows:
>>>
>>> 80.0 Mpccm/h 4.65320708035e+26 cm 1.97387969592 dimensionless
>>> 80.0 Mpccm/h 1.19455177073e+26 cm 1.97343368065 dimensionless
>>> 80.0 Mpccm/h 1.21507519065e+26 cm 1.97342592996 dimensionless
>>>
>>> Then I made the following change in cosmology.py:
>>>
>>> (root) ~/yt-conda/src/yt-hg/yt/utilities $diff cosmology.py
>>> cosmology.py0
>>> 111,112c111,112
>>> <         return self.quan((self.hubble_distance() *
>>> <                 trapzint(self.inverse_expansion_factor, z_i,
>>> z_f)).value, 'cmcm')
>>> ---
>>> >         return (self.hubble_distance() *
>>> >                 trapzint(self.inverse_expansion_factor, z_i,
>>> z_f)).in_base(self.unit_system)
>>>
>>>
>>> Then I get:
>>> 80.0 Mpccm/h 4.65320708035e+26 cmcm 1.92163908776 dimensionless
>>> 80.0 Mpccm/h 3.62771515661e+26 cmcm 1.92124702027 dimensionless
>>>
>>> With the change of unit from cm to cmcm, the light rays have the right
>>> span.
>>>
>>> Even though this solves my problem, I am not sure if similar problem
>>> still exists. For example, instead of making change in
>>> comoving_radial_distance, we might need to change hubble_distance into
>>> comoving units. Hopefully someone familiar with yt unit system could check
>>> this.
>>>
>>> Thanks,
>>> Pengfei
>>>
>>> _______________________________________________
>>> yt-users mailing list
>>> yt-users at lists.spacepope.org
>>> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>>>
>>>
>>
>> _______________________________________________
>> yt-users mailing list
>> yt-users at lists.spacepope.org
>> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>>
>>
>
> _______________________________________________
> yt-users mailing list
> yt-users at lists.spacepope.org
> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.spacepope.org/pipermail/yt-users-spacepope.org/attachments/20160809/0083c2c9/attachment.html>


More information about the yt-users mailing list