[enhancement] Add relativistic component to cosmology calculations #1714

brittonsmith · 2018-03-07T02:56:51Z

Update: I've finally come back to this and it should now be in a finished state.

This adds the relativistic component of the energy density of the universe, omega_radiation, to the cosmology calculations. Since there is no analytic solution to the Friedmann equations when omega_radiation > 0, we now numerically integrate using the trapezoid rule and then linearly interpolate in log(a) vs. log(t).

To get z_from_t, we reverse that table and interpolate. In this case, I also iteratively adjust the bounds of the log(a) table to be outside of the calculated log(a) values. This makes sure the integration is accurate and the ends of the table. This is most important at the high redshift end.

It's also important to note that the previous z_from_t and t_from_z calculations did not support all cosmologies, whereas this does.

In response to @jzuhone's comment about the hubble_time calculation, I've switched it to the correct formulation. @jzuhone was definitely right all along about that.

Differences of redshift/time conversions (t_from_z and z_from_t) are maximally about 1e-5. It's not nothing, but I think quite acceptable.

This is now supported in the Enzo code, so that frontend is updated to include it. I have added it as a standard dataset attribute for cosmology simulations.

jzuhone · 2018-03-07T19:50:39Z

So this is a bit tangential, but I hadn't really looked at the hubble_time method until now, and now I realize that what it actually computes is the age of the universe. When I hear "hubble time" I think of 1/H_0.

Would it make sense to address this naming issue in this PR as well? Maybe rename the hubble_time method to something about the age of the universe and then let hubble_time point to it with a deprecation warning?

Or am I missing something here entirely?

brittonsmith · 2018-03-07T22:07:10Z

I will use "hubble time" in a redshift-dependent way in speech, as in "the hubble time at redshift 20", but maybe that's just me. Does anyone else do this?

jzuhone · 2018-03-07T22:14:48Z

I've always seen it as "Hubble time" = t_H = 1/H_0. For a flat matter-only universe, the age of the universe is t = 2/3t_H = 2/(3H_0). But the hubble_time method is clearly calculating the age and not the "Hubble time".

jzuhone · 2018-03-07T22:15:45Z

Obviously "Hubble time" is redshift-dependent since H is a parameter which depends on z, but the method itself seems to compute not this but the age.

brittonsmith · 2018-03-08T18:07:31Z

Alright, what do you suggest then? Should hubble_time be deprecated in favor of something like universe_age?

jzuhone · 2018-03-08T18:23:03Z

I think that works, but open to suggestions (also open to challenges from anyone who thinks I am completely full of it)

ngoldbaum · 2018-03-08T18:26:22Z

Is it useful to also have a helper that returns 1/H? Not sure what that would be called if hubble_time is taken.

brittonsmith · 2018-03-08T18:47:41Z

I'm willing to say @jzuhone is correct, in which case 1/H should be hubble_time, but that does break things. I suppose we already have hubble_parameter, which returns H(z), so maybe it's not necessary and we should just deprecate hubble_time.

…lation from numerically integrated table for z_from_t and t_from_z.

…ribute for cosmological datasets.

remove table caching since it doesn't save that much time.

brittonsmith · 2019-02-04T20:00:30Z

I've finally gotten around to finishing this off. It should be ready for review now.

ngoldbaum

I haven't reviewed the math in the new a_from_t and t_from_a functions.

Do you think the existing tests for the cosmology calculator are sufficient to be confident about the changes here? It looks like only hubble_time,t_from_z, and z_from_t are explicitly tested. It also looks like hubble_time and t_from_z are tested to be identical in the default cosmology, which I guess is still the case.

Still would be nice to get some more tests, particularly for the new functionality you've added.

Do you have any Enzo datasets with nonzero omega_radiation we could add to yt-project.org/data (if we don't have those already). That would make it easier to debug any issues people report in the future about this. Could also use such a dataset for a test of this functionality.

ngoldbaum · 2019-02-04T20:15:35Z

yt/utilities/cosmology.py

@@ -319,9 +323,9 @@ def lookback_time(self, z_i, z_f):
        return (trapzint(self.age_integrand, z_i, z_f) / \
                self.hubble_constant).in_base(self.unit_system)

-    def hubble_time(self, z, z_inf=1e6):


I realize we talked about this already, but would you be ok with keeping the old calculation but deprecating the name hubble_time. You could then add a new function which returns one over the hubble constant. I worry about users who will now get different results from this function in the next yt version with no warning. If you feel that's not a concern worth worrying about I'd love to hear why you think so.

Sure, I can do that. I think the counterargument is that this could be seen as constituting a bugfix, but since the correct behavior is basically trivial, I can just include that in the deprecation warning.

ngoldbaum · 2019-02-04T20:16:42Z

yt/utilities/cosmology.py

@@ -388,10 +384,11 @@ def hubble_parameter(self, z):
        >>> print(co.hubble_parameter(1.0).in_units("km/s/Mpc"))

        """
-        return self.hubble_constant.in_base(self.unit_system) * self.expansion_factor(z)
+        return self.hubble_constant.in_base(self.unit_system) * \
+          self.expansion_factor(z)


Minor nit, can you format this using parens to avoid the continuation character:

return (self.hubble_constant.in_base(self.unit_system) * self.expansion_factor(z))

…cter.

brittonsmith · 2019-02-04T22:45:28Z

@ngoldbaum, I'm happy to add some tests of each function if I'm allowed to just compare with some hard-coded answers. I can also do some comparisons with the old analytic functions from Enzo for the supported cosmologies. I'll also see what I can do about some Enzo data.

…est.

brittonsmith · 2019-02-07T01:06:27Z

I've issues a PR to the website repo to add an Enzo dataset with non-zero omega_radiation. I've also added some small tests using this dataset here.

ngoldbaum · 2019-02-07T14:56:43Z

@yt-fido test this please

brittonsmith · 2019-02-07T19:38:46Z

These test failures are real. I'll look into them.

…n message.

…ion classes.

brittonsmith · 2019-02-07T22:47:01Z

The perils of giving keyword arguments as positional arguments...

…to standard attributes.

brittonsmith · 2019-02-08T02:41:21Z

The rest of the test failures likely result from minor changes in cosmology calculations. I'll investigate and update when I have the chance.

ngoldbaum · 2019-02-08T03:43:21Z

It looks like four of the cookbook recipes are raising errors on python2.

I think bumping the answer numbers is warranted if the amswers are sensitive to that and the changes in answers are small.

brittonsmith · 2019-02-09T21:12:20Z

@yt-fido test this please

brittonsmith · 2019-02-10T18:11:48Z

I traced the answer test failures down to a change that I never intended to keep. That's fixed and they now all pass without having to update answers. The only remaining failures are due to pyaml not being installed for the minimal dependencies tests. Would it be ok to add that to those builds?

Edit: I meant pyyaml, not pyaml. They are different.

ngoldbaum · 2019-02-10T18:38:29Z

So the whole point of the minimal dependencies tests is that the test suite runs without errors with a minimal set of dependencies installed. Can you make it so the tests that depend on pyyaml only run if it’s importable?

ngoldbaum · 2019-02-10T18:39:23Z

But definitely add it to test_requirements.txt.

ngoldbaum · 2019-02-11T19:38:11Z

Awesome, thanks so much for sticking with it on this one! I'm liking all the new tests :)

brittonsmith changed the title ~~[enhancement] Add relativistic component to cosmology calculations~~ [wip, enhancement] Add relativistic component to cosmology calculations Apr 10, 2018

brittonsmith mentioned this pull request Jun 8, 2018

Add support for particle fields and cosmology in Enzo-P frontend #1490

Merged

brittonsmith added 4 commits February 3, 2019 20:39

Adding omega_radiation term to expansion factor and implement interpo…

09a9698

…lation from numerically integrated table for z_from_t and t_from_z.

Add support for omega_radiation to Enzo frontend and add it as an att…

5a44545

…ribute for cosmological datasets.

Fix hubble time.

0a34a91

Correcting cosmology calculator docs.

2f5921e

brittonsmith force-pushed the omegarad branch from 4f09460 to 2f5921e Compare February 4, 2019 04:39

brittonsmith added 2 commits February 4, 2019 11:13

Remove IntegralTable class and use a function instead. Also,

b8dbebe

remove table caching since it doesn't save that much time.

Use iteration to widen integration bins for a_from_t calculation.

fa25e05

brittonsmith changed the title ~~[wip, enhancement] Add relativistic component to cosmology calculations~~ [enhancement] Add relativistic component to cosmology calculations Feb 4, 2019

brittonsmith requested review from ngoldbaum and jzuhone February 4, 2019 19:59

ngoldbaum reviewed Feb 4, 2019

View reviewed changes

Add deprecation warning for hubble_time and remove continuation chara…

ce06557

…cter.

brittonsmith added 6 commits February 4, 2019 15:09

Remove hubble_time test and widen redshift range for t/z conversion t…

c9d3a9b

…est.

Add test to compare cosmology z/t conversion with analytic expressions.

982d859

Fix line appearance.

a25035c

Add small tolerance to comoving_volume calculation.

ea06733

Add cosmology answer tests.

6d63545

Add dark factor test.

9dc2ee4

Make sure nbins argument to linspace is an int.

e423e92

yt-project deleted a comment from codecov bot Feb 6, 2019

Add non-zero omega_radiation test.

a69161e

brittonsmith added 5 commits February 7, 2019 14:20

Return hubble_time function to original limits and clarify deprecatio…

06dfcf1

…n message.

Change kwarg order.

b0b365d

Fix all instantiations of Cosmology objects to explicitly use keywords.

366864e

Add support for omega_radiation term to EnzoCosmology and EnzoSimulat…

e3e6a8f

…ion classes.

Remove hubble_time from docs.

ff8cd93

Set default value for omega_radiation in EnzoSimulation and add term …

918cadb

…to standard attributes.

brittonsmith added 3 commits February 8, 2019 14:45

Ensure float arithmetic for python 2.

8a416f6

Return bins default to original value.

de9caf2

Add cosmology answer tests with data in yaml file.

2d09839

brittonsmith force-pushed the omegarad branch from bed2f66 to 2d09839 Compare February 9, 2019 21:10

brittonsmith added 2 commits February 9, 2019 13:20

Ensure a few more floats for python 2 division.

9912ffa

Return yaml import.

ea88c83

brittonsmith added 2 commits February 10, 2019 12:49

Add yaml on demand import.

a1a0120

Add pyyaml to test requirements.

c847ee2

ngoldbaum merged commit 7f0b1de into yt-project:master Feb 11, 2019

brittonsmith deleted the omegarad branch February 11, 2019 20:07

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[enhancement] Add relativistic component to cosmology calculations #1714

[enhancement] Add relativistic component to cosmology calculations #1714

brittonsmith commented Mar 7, 2018 •

edited

jzuhone commented Mar 7, 2018

brittonsmith commented Mar 7, 2018

jzuhone commented Mar 7, 2018

jzuhone commented Mar 7, 2018

brittonsmith commented Mar 8, 2018

jzuhone commented Mar 8, 2018

ngoldbaum commented Mar 8, 2018 •

edited

brittonsmith commented Mar 8, 2018

brittonsmith commented Feb 4, 2019

ngoldbaum left a comment

ngoldbaum Feb 4, 2019

brittonsmith Feb 4, 2019

ngoldbaum Feb 4, 2019

brittonsmith commented Feb 4, 2019

brittonsmith commented Feb 7, 2019

ngoldbaum commented Feb 7, 2019

brittonsmith commented Feb 7, 2019

brittonsmith commented Feb 7, 2019

brittonsmith commented Feb 8, 2019

ngoldbaum commented Feb 8, 2019

brittonsmith commented Feb 9, 2019

brittonsmith commented Feb 10, 2019 •

edited

ngoldbaum commented Feb 10, 2019

ngoldbaum commented Feb 10, 2019

ngoldbaum commented Feb 11, 2019

[enhancement] Add relativistic component to cosmology calculations #1714

[enhancement] Add relativistic component to cosmology calculations #1714

Conversation

brittonsmith commented Mar 7, 2018 • edited

jzuhone commented Mar 7, 2018

brittonsmith commented Mar 7, 2018

jzuhone commented Mar 7, 2018

jzuhone commented Mar 7, 2018

brittonsmith commented Mar 8, 2018

jzuhone commented Mar 8, 2018

ngoldbaum commented Mar 8, 2018 • edited

brittonsmith commented Mar 8, 2018

brittonsmith commented Feb 4, 2019

ngoldbaum left a comment

Choose a reason for hiding this comment

ngoldbaum Feb 4, 2019

Choose a reason for hiding this comment

brittonsmith Feb 4, 2019

Choose a reason for hiding this comment

ngoldbaum Feb 4, 2019

Choose a reason for hiding this comment

brittonsmith commented Feb 4, 2019

brittonsmith commented Feb 7, 2019

ngoldbaum commented Feb 7, 2019

brittonsmith commented Feb 7, 2019

brittonsmith commented Feb 7, 2019

brittonsmith commented Feb 8, 2019

ngoldbaum commented Feb 8, 2019

brittonsmith commented Feb 9, 2019

brittonsmith commented Feb 10, 2019 • edited

ngoldbaum commented Feb 10, 2019

ngoldbaum commented Feb 10, 2019

ngoldbaum commented Feb 11, 2019

brittonsmith commented Mar 7, 2018 •

edited

ngoldbaum commented Mar 8, 2018 •

edited

brittonsmith commented Feb 10, 2019 •

edited