Re-implementation of ATLAS prompt photon production

[jacoterh]

This PR reimplements ATLAS prompt photon production at 13 Tev. The validphys report is linked here.

observables = ["XSEC"]

for obs in observables:

    new_implementation = f"ATLAS_PH_13TEV_{obs}"
    old_implementation = f"ATLAS_PH_13TEV_{obs}"

    inp1 = {
        "dataset_input": {"dataset": f"{new_implementation}"},
        "theoryid": 200,
        "use_cuts": "internal",
        "t0pdfset": "NNPDF40_nnlo_as_01180",
        "use_t0": True,
    }
    inp2 = {
        "dataset_input": {"dataset": f"{old_implementation}", "variant": "legacy"},
        "theoryid": 200,
        "use_cuts": "internal",
        "t0pdfset": "NNPDF40_nnlo_as_01180",
        "use_t0": True,
    }

    covmat1 = API.covmat_from_systematics(**inp1)
    covmat2 = API.covmat_from_systematics(**inp2)

    t0_covmat1 = API.t0_covmat_from_systematics(**inp1)
    t0_covmat2 = API.t0_covmat_from_systematics(**inp2)

    print(f"Comparison for {new_implementation}")
    print(np.all(np.isclose(covmat1, covmat2)))
    print(np.all(np.isclose(t0_covmat1, t0_covmat2)))

gives

Comparison for ATLAS_PH_13TEV_XSEC
False
False

The ratio of the eigenvalues of the legacy and the new covmat is close to unity though. A summary of what changed:

• k1 (the photon pseudo rapidity) in kinematics_XSEC was incorrect. It did not correspond to the midpoint of each bin, but to half the bin width.
• All systematics reported on HEPdata are now taken into account. Before there were only two, now we have 91. This obviously affects the central value (through shifts due to asymmetric uncertainties) and the covmat itself.

[scarlehoff]

Instead of ET^2, I think it makes more sense to have ET in the kinematics (and change the process_options.py accordingly).

I'm also not sure the value of x is computed correctly, since we should have x1 and x2.

Also, since you have the rapidity of the photon I think it makes more sense to use that for to access the x1/x2 kinematics? (because Et/s at best is giving you something like that x1*x2)

I'm also a bit worried about the shifts in the central data:

2nd and 4th plots in master:

2nd and 4th plots in this branch:

Could you double-check the values of eta? Are you sure that the data is in the exactly same order as before?

[jacoterh]

Hi @scarlehoff , I should have fixed the x Q2 mapping, please check. The updated vp report can be found here. Regarding the pseudo rapidity on master, it did not correspond to the midpoint of each bin, but to half the bin width. The new vp report looks fine to me. Let me know whether you agree.

By the way, I don't find any old implementation of ATLAS_PH_8TEV_XSEC even though this dataset is listed on the wiki. What's up with that?

[scarlehoff]

By the way, I don't find any old implementation of ATLAS_PH_8TEV_XSEC even though this dataset is listed on the wiki. What's up with that?

It was not in 4.0 so I didn't port it, but it is probably in the old buildmaster. The nice thing is that there's no much check to do for those ^^ (just that the results are not too crazy ofc).
If you want I can start by looking at the 13 TeV one and merge it when/if done

[jacoterh]

I see, then let's try to just merge this and work on the 8TeV one later. It's ready for review in any case!

[scarlehoff]

lgtm, please update also the version / version comment here

[jacoterh]

Done, I simply forgot, thanks for pointing it out