SciPost logo

SciPost Submission Page

Illuminating the photon content of the proton within a global PDF analysis

by Valerio Bertone, Stefano Carrazza, Nathan P. Hartland, Juan Rojo

This is not the latest submitted version.

This Submission thread is now published as

Submission summary

Authors (as Contributors): Stefano Carrazza
Submission information
Arxiv Link: (pdf)
Date submitted: 2018-05-01 02:00
Submitted by: Carrazza, Stefano
Submitted to: SciPost Physics
Ontological classification
Academic field: Physics
  • High-Energy Physics - Experiment
  • High-Energy Physics - Phenomenology
  • Nuclear Physics - Theory
Approach: Theoretical


Precision phenomenology at the LHC requires accounting for both higher-order QCD and electroweak corrections as well as for photon-initiated subprocesses. Building upon the recent NNPDF3.1 fit, in this work the photon content of the proton is determined within a global analysis supplemented by the LUXqed constraint relating the photon PDF to lepton-proton scattering structure functions: NNPDF3.1luxQED. The uncertainties on the resulting photon PDF are at the level of a few percent, with photons carrying up to 0.5% of the proton's momentum. We study the phenomenological implications of NNPDF3.1luxQED at the LHC for Drell-Yan, vector boson pair, top quark pair, and Higgs plus vector boson production. We find that photon-initiated contributions can be significant for many processes, leading to corrections of up to 20%. Our results represent a state-of-the-art determination of the partonic structure of the proton including its photon component.

Current status:
Has been resubmitted

Author comments upon resubmission

Updated version containing referee's suggestions as listed in the respective replies.

Reports on this Submission

Anonymous Report 3 on 2018-5-11 (Invited Report)

  • Cite as: Anonymous, Report on arXiv:1712.07053v2, delivered 2018-05-11, doi: 10.21468/SciPost.Report.445


As stated in my original report.


There remains an issue with the comparison between the 3.0 and new Lux set which as far as I can see has not been resolved, while consistency between these two is still claimed.


Firstly, I would like to apologise for the long delay in my responding. I did not receive a notification from the journal of your reply, and so missed this for some time. I would have replied much earlier otherwise.

I would like to thank the authors for their careful response to my comments. However, I feel I have to insist upon this issue of the NNPDF3.0 vs. 3.1LUXqed comparison before I can personally recommend publication.

Two reasons are given for this difference. The first reason, namely the impact of higher-order corrections in the evolution, is clearly quite small and certainly (as the authors acknowledge) far too small to explain the size of the difference. This leaves the second stated reason, namely the fact that the input is due to the 2.3 fit, which used a different form of evolution to the current unified one. In other words, the claim is that if the 2.3 fit were redone, but using the unified (3.0) evolution, the NNPDF3.0 and 3.1LUXqed would (at least largely) converge. I have two objections to this:

1) Even if we accept the logic as described, this only remains a possible reason, as no such 2.3 fit, but now with unified evolution, has been performed. Perhaps if this were done the discrepancy would remain; from the information we have, we (at best) simply do not know. I am not suggesting this is done (that would clearly be beyond the scope of the paper), but nonetheless there is still no reliable basis to claim, as is still currently done in the introduction and conclusion, that the updated PDF set is fully consistent within uncertainties with NNPDF3.0QED.

2) My stronger objection is that as far as I can tell the logic described by the authors would actually lead to an increase in the tension between NNPDF3.0 vs. 3.1LUXqed. In particular, as the unified evolution is faster at low x in comparison to that used in the 2.3 set, if this fit were to be repeated using the unified evolution, we would surely require a *smaller* and not a larger, input photon at low x to fit the same data, i.e. to compensate the faster evolution. In other words, the tension would if anything increase.

Given this, it appears to me that for whatever reason the previous NNPDF agnostic QED fits do not give results at low x which are consistent with 3.1LUXqed, for a reason which cannot be due to this differing evolution (or at the very least has not been shown to be), but rather something relating to the earlier fit itself. The reason for this remains unclear, but consistency definitely cannot be claimed.

I look forward to receiving further comments on this, if I am mistaken; though even if I am in terms of point (2), point (1) remains. So, the comments claiming consistency between 3.0 and 3.1LUXqed should be rephrased - any consistency cannot currently be claimed - and the discussion around Fig. 3.3 updated. Ideally reasons for this apparently genuine disagreement would be considered, but I appreciate at this stage this may not realistic.

Requested changes

As described above.

  • validity: good
  • significance: top
  • originality: good
  • clarity: high
  • formatting: excellent
  • grammar: excellent

Anonymous on 2018-05-24  [id 257]

(in reply to Report 3 on 2018-05-11)
answer to question
reply to objection

First of all, we fully agree with the referee that we cannot really claim consistency between NNPDF3.0QED and NNPDF3.1LUXqed in the small-x region below roughly x=0.01. Irrespective of the reason, this is just an empirical fact that the two sets do not agree within the quoted uncertainties in this region. We have modified the text to reflect better this fact, which perhaps was not that clearly stated in the previous revision of the paper. We have also added a remark that for the region more relevant for LHC phenomenology, namely x > 0.01 (which roughly corresponds to Q > MZ), both sets agree indeed within uncertainties.

Concerning point 1), we have toned down the statements concerning the possible explanation for the discrepancy between NNPDF3.0 and NNPDF3.1luxQED. As the referee rightly points out, without redoing the same fit with the correct evolution settings (and comparing exactly like with like) it is not really possible to categorically conclude that this is the dominant reason for the differences. So now we state that this is a possible reason (certainly it does contribute to the differences) but we don't know for sure that this is the main one, since one could conceive other options, due to the radically different methodologies adopted in the two sets. As the referee acknowledges, a final clarification of this point would be beyond the scope of this paper.

Concerning point 2), it seems to us that the situation is a bit more complex than that. Indeed, the NNPDF2.3QED photon PDF was constrained mostly by Drell-Yan data at the LHC (since the constraints only from DIS structure functions were extremely loose). Therefore the right comparison would be at the level of PDF luminosities, rather than at the level of PDFs themselves. Note for example that in Fig. 3 of arXiv:1606.07130 the effects of the 3.0QED evolution actually lead to a small suppression of the photon PDF at large-x rather than an enhancement. This would be relevant for say the LHCb low-mass Drell-Yan data used in the NNPDF2.3QED fit, where one parton is at low-x and the other at small-x. So one should look at the kinematics of all hadronic datasets used in NNPDF2.3QED point by point to see how these differences in the evolution feed into the resulting photon, and this is rather complicated without doing the full fit. All in all, the only reason to conclusively set this point would be to redo NNPDF2.3QED with the same evolution as NNPDF3.0QED.

In any case, as mentioned above, we agree with the referee that one cannot exclude other sources of difference between NNPDF2.3/3.0QED and NNPDF3.1luxQED, perhaps of methodological origin. We have made sure that this fact is properly reflected in the text so that there are no ambiguities.

We hope that with the modifications that we have carried out in the new revision of the paper, the referee will consider it suitable for publication in SciPost.

Anonymous on 2018-05-28  [id 261]

(in reply to Anonymous Comment on 2018-05-24 [id 257])

Thank you for this update, I think this describes the situation more clearly. As you say, empirically one cannot claim 'consistency' in the sense that the PDF sets do indeed not agree for all x regions.

The remaining question is whether the sets are 'consistent' in the sense that, if the same theoretical framework for the evolution were used throughout, the results would be consistent. The point about the suppression at high x for the 3.0 (vs. 2.3) evolution is certainly a good one, although I personally still find it very hard to see how this sort of effect, which is clearly relatively mild, could lead to such a sizeable suppression in the 3.0QED photon at low x. Somehow it seems that there is something else methodological in the original 2.3QED case (after all a reweighting to a rather unconstraining dataset as we know) at play here. I may be wrong though.

In any case as we are all agreed this is beyond the scope of the current paper. I am completely happy with the updated version and offer no further comments or suggestions.

Login to report or comment