Implementing ATLAS_PH-8TEV_XSEC

[jacoterh]

This PR implements ATLAS prompt photon production at 8 TeV, similar to #2209 at 13 TeV. No legacy implementation exists for this dataset as it wasn't ported from buildmaster.

I tried to produce a validphys report, but it complains it cannot use an old theory with a purely new dataset:

[ERROR]: Bad configuration encountered: Cannot use an old theory with a purely new dataset (ATLAS_PH_8TEV_XSEC)

I used this card.

[jacoterh]

Hi @RoyStegeman , I forgot how we agreed to proceed here at the last code meeting. Do you remember it? Sorry!

[RoyStegeman]

It was suggested to transform the APPLgrid to pineappl format and add it to the new theory (708)

[jacoterh]

Hi @enocera, I noticed that the old buildmaster implementation of this dataset doesn't provide the applgrid for the rapidity region 1.81<|eta(gamma)| < 2.37. It isn't listed at https://github.com/NNPDF/applgrids/tree/master/ATLASPHT12, while the 13 TeV counterpart does provide the applgrid in this region. Do you recall the reason for this? The hepdata entry is https://www.hepdata.net/record/ins1457605?version=1&table=Table%204 for reference.

[enocera]

Dear @jacoterh as far as I can recollect, the reason is historical. Prompt photon data (8 TeV and 13 teV) were originally studied in https://arxiv.org/pdf/1802.03021. The NNPDF4.0 implementation closely follows the results in that paper. On page 4, one can read: "the most forward rapidity bin, 1.81 ≤ |ηγ | < 2.37, is excluded from the fit due to the tensions between the experimental data and the theoretical predictions." I guess that, for this reason, the PineAPPL grids were not generated at all in NNPDF4.0. At 13 TeV, the paper does not report any restriction.

[enocera]

For now, I suggest to assume that this statement (and the NNPDF4.0 selection) hold true.

[scarlehoff]

@jacoterh note that it is possible to shift a bin for a number of points like in this case:

nnpdf/nnpdf_data/nnpdf_data/commondata/ATLAS_Z0J_8TEV/metadata.yaml

Line 107 in d7bfbd0

shifts:

In this example ATLASZPT8TEVMDIST-ATLASZPT8TEV-MLLBIN4_ptZ and ATLASZPT8TEVMDIST-ATLASZPT8TEV-MLLBIN6_ptZ both get shifted by 10 to account for the gap between the data and the theory.

[jacoterh]

Thanks @enocera, that clarifies it!

[jacoterh]

FK tables have been added to theory 708. See the below link for the data theory comparison with the new commondata implementation: https://vp.nnpdf.science/RyQDSR8VRSONW9yLre6EzA==/

The grids excluded the first 4 ET bins in each rapidity bin, this shift has been accounted for in the filter as well.

[scarlehoff]

Given the shape of the ET distribution, @enocera, do we want to cut a few more bins in ET? The 13 TeV distribution looks a bit better but it might be because of EWK k-factors? (although they are quite small...)

[scarlehoff]

I haven't been paying much attention to the filter files themselves, but won't many of the functions be similar/equal to those from ATLAS_PH_13TEV?

[jacoterh]

Yes, the filter is equivalent to ATLAS_PH_13TEV, that's right

[scarlehoff]

Will you merge both or do you want to keep it explicitly separted?