Running of alpha_em in `runcards.update` about eko

[felixhekhorn]

@niclaurenti how much is this a problem? i.e. what do you do in real life? I guess the default is no running ... in short: should we worry or no?

[alecandido]

In principle, this should not be used during actual PDF evolution: we decided that $\alpha$ evolution would have been inconsistent, and so we do not explicitly support it.
But since it has been implemented, we provide it as a library function. I see no issue in not exposing it through the runcard, since we are actively deciding to never use it.

[felixhekhorn]

Even fine to just close again ... since I touched that part in #238, I just wanted to mention it ...

[alecandido]

Let's wait for @niclaurenti acknowledgement, and then we can close.

[niclaurenti]

@niclaurenti how much is this a problem? i.e. what do you do in real life? I guess the default is no running ... in short: should we worry or no?

In this moment I'm not using the running of alphaem so it will not be a mess. I don't know if in the future will use it

[niclaurenti]

But I did't understand exactly why the running of alphaem is not accessible now

[felixhekhorn]

But I did't understand exactly why the running of alphaem is not accessible now

Only from the old-style theory runcards - because here

eko/src/eko/io/runcards.py

Lines 291 to 297 in 8f0e3a7

	new["couplings"] = dict(
	alphas=old["alphas"],
	alphaem=alphaem,
	scale=old["Qref"],
	num_flavs_ref=old["nfref"],
	max_num_flavs=old["MaxNfAs"],
	)

I'm not setting the em_running property:

eko/src/eko/quantities/couplings.py

Line 32 in 8f0e3a7

em_running: bool = False

(which by the way breaks again our no-default-rule - but, as already said, we can not keep that up most likely due to #145 )

[alecandido]

In this moment I'm not using the running of alphaem so it will not be a mess. I don't know if in the future will use it

Wait... don't you know? @niclaurenti

This was the beginning of the story, 1y ago, when we asked Christopher, Alessandro Vicini, and friends whether $\alpha$ was used running in the partonic cross-sections calculations, and they replied "never".

Fine to allow people to get the $\alpha$ evolution stand-alone, but which would be the context in which you know you will have to use a running electromagnetic coupling during PDF evolution? Or is it just a fully-hypothetical statement, with no actual use case currently known?

[niclaurenti]

In this moment I'm not using the running of alphaem so it will not be a mess. I don't know if in the future will use it

Wait... don't you know? @niclaurenti

This was the beginning of the story, 1y ago, when we asked Christopher, Alessandro Vicini, and friends whether α was used running in the partonic cross-sections calculations, and they replied "never".

Fine to allow people to get the α evolution stand-alone, but which would be the context in which you know you will have to use a running electromagnetic coupling during PDF evolution? Or is it just a fully-hypothetical statement, with no actual use case currently known?

Ah ok. Sorry I misunderstood the meaning of the issue. I thought that it was no longer possible to use the alphaem running in eko.
I know that in the grids aem is fixed.

However, the reason I implemented it was mainly for benchmarking against apfel (I had a difference and I wanted to understand whether it was due to a bug or due to the running of alpha)

[alecandido]

However, the reason I implemented it was mainly for benchmarking against apfel (I had a difference and I wanted to understand whether it was due to a bug or due to the running of alpha)

This could have been reasonable, but, by various experiments, we know the impact to be negligible, and the effort to support it to be much less negligible.

The new runcards format even allows for a running coupling, but it is a potential future feature, not yet supported (@felixhekhorn we should consider setting it to False by default, since it is not currently a real option).

I believe that this exhausts the discussion of this issue, thus I'm closing it.
If I'm wrong, feel free to reopen it :)