Comments as received on draft for paper presentation ----------------------------------------------------- Achim Geiser: -------------- Thanks a lot for the great paper draft and the nice results. I leave simple style comments to others, so please find my significant comments below: section 4: The "experimental" uncertainty is now even less experimental than it was before, since the hadronization uncertainties are now also included. In some ZEUS papers we have just called it "fit" uncertainty, but I know that was not acceptable to H1. So, my compromise proposal (similar to H1/ZEUS paper arXiv:1804.01019) line 231: "exp/fit" line 232: "where exp/fit denotes the fit uncertainty based on the experimental input uncertainties and the hadronization uncertainty" line 234: "the fit uncertainty" line 251: "The similarity ... no additional tension." I don't think this is true. Similar chis/dof > 1 indicates similar tension, also in the jet data. Thus rather: "The similarity of the chi2/d.o.f. values indicates a similar level of tension for the jet and inclusive data." or "The similarity of the chi2/d.o.f. values indicates that the addition of the jet data does not change the average level of tension in the data." or "The similarity of the chi2/d.o.f. values indicates that the addition of the jet data does not increase the average level of tension in the data." line 259: "the negative gluon term" -> "the flexible gluon term" There is nothing in this term that forces the gluon to become negative, nor can the gluon be prevented to become negative (at low enough scales) even without this term. A more flexible parametrisation can never bias the result, as long as the fit converges properly. A more restrictive one can. Thus "could bias" -> "could significantly alter". line 266: "the details" -> "these details". It indicates nothing about potential other variations of the gluon parametrisation. line 278: I would move this difference right to the beginning of the section, since it is a basic conceptual difference that is the main asset of this paper w.r.t. the others. The difference in the treatment of the scale uncertainties in the previous paragraph, although numerically more important, is only a technical difference and should not be highlighted as if it were a conceptual one. Also, it could be stated that the 100% scale correlation option is the "traditional" variant. I am personally convinced the treatement we use use here is better than the traditional one, so I do *not* propose (and would strongly oppose) changing our default treatment. Section6: Repeat here (or elsewhere) that the jets were obtained with the kt algorithm and R=1. Remind that it was established (cite arXiv:1003.2923) that at HERA, using the kt or anti-kt algorithms (as used at LHC) is qualitatively equivalent. Figures 9-12: It is hard to see the comparison on log scale only. Providing ratio subpanels would be helpful. line 336: "scale uncertainties were not considered for the comparisons to data". I don't think this is acceptable. The NNLO predictions have significant scale uncertainties (from the jet matrix elements), and these should be shown (separately). This is different from the potential scale uncertanties of the PDFs that were discussed in the previous sections, which can indeed not be usefully quantified. Best regards, Achim (from vacation) - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - Mandy: Wrt to Achim's comment on scale uncertainties. He wants them on the data vs theory figures. The question is then exactly what does he want? ie the theory in the figures is for fixed alphas=0.1155 and, so does he want us i) to take that fit with fixed parameters and just alter the scale (as I did in evaluating where to put the \mu cut) and look at the predictions or does he want ii)refits with the new scale (which will not be good if alphas=0.1155 as we have already seen) and predictions from the refit? I think that will show little change just as the PDFs showed little change. I think he must want i). But we have never shown i) in previous work. We need to clarify with him, but I would say there is no hurry. ______________________________________________________________________________ Aharon Levy: Dear Editorial Board members, Congratulations on completing the HERAPDF2.0 family. Well done!! Here are some comments to a very well written paper. I tried to read it from the point of view of someone not an expert on the subject. minor comments: l. 11, ….order (NNLO). l. 34, ….formalism using the … l. 53, …established [reference]. l. 91, …equation (32) l. 93, …fit (see…below). l. 111, …running electromagnetic… l. 234, remove one ’the’. l. 241, Eq. (7). The (7) did not come out in line 231. significant(?): l. 37, remove sentence 'An analysis….family.' l. 51, remove 'the analysis…at NNLO.' This belongs to the conclusions. l. 73, ’several reasons’ - are they all listed in this paragraph? l. 81, In Table I, ’normalised’ appears in the data set column and in the caption. Is it obvious what is meant by normalised? footnote 3, Is it clear why the value of 25 was chosen. All it says is the it should be > 15. l. 203, what is meant by ’technical reasons’? l. 210, when trying to make a point about precision, why use ~ for the uncertainty? Section 3.4 is titled ‘hadronisation uncertainties’ so is the number given in l. 215 the one that was used here ? It appears with ref [2], meaning it is probably the one used in that paper. What is the number used for the present analysis? lines 340, 360 …ZEUS and H1…Throughout the paper the order is alphabetic, namely H1 and ZEUS. I thought it would make sense to keep this order. l. 356, is this last sentence the highlight of the paper? It would be nice to end this with a more significant statement. References: Halina’s name is mispelled in refs [2] and [21].—> Abramowicz. refs [11] and [12] have the same arXiv number. Change ref [12] to hep-ph/0609285 ref [17] V. Andreev et al. Figures: Fig. 2: a) not good choice of the light blue color (alpha_S free fit). Is that the line passing through the blue points? b) red points connected with a dark line. c) red points connected with a red line. Why is the x-axis scale of a) different from b) and c)? Figs. 3, 4 and 7 the sea label in the figure and the captions are different. Best regards, Aharon _______________________________________________________________________________ Erich Lohrmann: Dear Colleagues, one of the very important results of the paper is the measurement of alpha_s at NNLO. It should find its proper place in the title of the paper and the abstract. Suggest for title: 'Impact of jet production data at NNLO on the determination of HERAPDF2.0 parton distributions and an alpha_s(M_z)' In the abstract giving alpha_s without the most important uncertainty is useless. Suggest: '..is alpha_s(M_Z)=0.1155+-0.0010(exp)+-0.0002(model) +0.0026/-0.0024 (scale).Scale uncertainties are discussed in detail...' Congratulations on an impressive paper. Best regards, Erich Lohrmann. _______________________________________________________________________________ Ewald Paul: thank you to all who have made possible this nice finalpaper on HERA PDFs". I have a comment (1) and asuggestion (2). (1) I think we should not talk about "similarities" in lines 22 and line 354, since we can talk about consistancies. (2) The lines 50 bto 54 suggest a useful phrasing of the main result: "The PDFs of previous studies and the new study including jets are consitent throughout the whole kinematic range." "The analysis presented here demonstrates concistancy of the jet data with the inclusive data on NNLO level." My suggestion: we talk about consistancy instead of simularities and we improve abstract and summary in the spirit of the nicely summerizing phrasing in the introduction. Best wishes for the paper, Ewald _______________________________________________________________________________ Masahiro Kuze: Beautiful analysis and paper. Comments are all minor: L15: does the phrase 'combined PDF and QCD fits' make a rigorous scientific sense? Are there different things such as 'PDF fit' and 'QCD fit' and they are combined? L111: electro magnetic -> electromagnetic L126: sorry, I cannot count 10 free parameters - A, B, C for five PDFs make 15? I guess some readers may think the same, so some explanation is needed. L228: (2.0 mu_r, 1.0 mu_r) -> the latter should be mu_f ? L241: I cannot find Eq. 7. L321: As discussed 'in' Section 3.3 L333: 'was entered' sounds colloquial. "each jet P_T entered in the cross section (calculation)" Figure 3,4: the plot shows xS, but the caption says x\Sigma Figure 6 caption says 0.116 but should be 0.1155 Figure 9, 11: label of the vertical axis (what is plotted) differs from what is said in the caption (I mean the notation) while they match in Figure 10,12. L438: at the same time -> At the same time Notations (I thought Brian is in the EB?): - s in alpha_s is a label (strong) and not a variable, so should in roman (it seems H1 papers make it in roman and ZEUS papers have it in italics) - T in p_T (as well as T in E_T, L332) is a label (Transverse) and not a variable, so should be in roman - v in u_v and d_v is a label (valence) and not a variable, ... (ditto) - min in Q^2_{min} ... (ditto) - f in mu_f in Figure 5 caption ... (ditto) _______________________________________________________________________ Peter Bussey Line 28 First sentence is a little cumbersome - I suggest to insert "recorded" before "at" and add a comma after HERA. Perhaps better: Data from....,ep, have been central.....[1]. Such data have been recorded at....at HERA. Then run on to continue with contents of second paragraph. 33 These analyses were based.... [avoid repetition] 34 I think you should not assume familiarity but state the order that these fits were performed so that the readers know exactly where things stood before this analysis. 36 Define NNLO .... inclusive DIS data [2] 37 Omit sentence, no need to make excuses. 38 The present analysis has been made possible by the recent provision of jet cross section predictions for ep scattering at NNLO by the NNLOJET authors and their collaborators. [be positive!] 41 to be constructed from -> to originate from 42 Can you say something more precise than "treated"? 46 omit "both" 49 All -> The [repetition with "entirely"] I think I would again say "inclusive DIS" 50 Omit "highly" - it's consistent or it is not. The sentence really needs a phrase of amplification - why should it not be consistent? Why is it a single data set and not two? 51 Omit "very" 53 Reference for this statement. I would omit sentence "It is..." and rephrase a bit. "With the assumption that it is also valid in hadron-hadron physics, PDF fits can be made to LHC data. However, they can be biased by any presence of physics beyond the Standard Model (BSM), including this tacitly in the fitted PDF values and thereby reducing the sensitivity to explicit searches for BSM physics owing to biased background predictions." 58 Omit final sentence, not needed. 62 coherent? Above it was "consistent". I'd stick with that but say what it means. 67 ... squared in the DIS process, This sentence has "published" three times! Omit "which were...[27]" as it is just background information here and makes the sentence cumbersome. 71 Omit ", however, only" [Basically, the reader just needs the facts.] Since the reader may now be a little confused as to what "new" means, I would repeat the reference or else mention the present analysis. 73 Say:... were excluded in order to ensure convergence of the perturbative series [EXPLAIN THIS!] and to ensure that the NNLO scale uncertainties were no greater tha 10%..... I think the rest of the text here is too chatty and you need to stick to the basic points. 76 $b$ 78 "bin" is jargon, should we say "interval" ? [passim] 78-9,82 had to be excluded -> were excluded [passim, too chatty] 80 were judged to be -> were [avoid subjectivity] 89 The present analysis was performed in the same way as the previous [more concise] 90 Does this mean "cros-section data with Q^2 > 3.5 GeV^2"? If so, say it like that; if not, make the meaning clearer? 91 I think you should repeat the crucial chisq definition here. 92 State what you mean by evolution and mention DGLAP (again). 97 Can you reference the second program?? 100 As explained in the next section? Better say that. 103 ... was made possible by the.... [sounds more forceful!]. Repeat the reference. 116 results 119 Could just say "For the gluon PDF, an additional term..." 123 I think this sentence would be better at the start of the section. 132-137 I think these lines should go before the sentence starting line 125. However it becomes a little unclear at what point the addition of new parameters stopped. If this means when all the parameters in 132-137 were used, and this was optimal, then I'd say "... could not be improved by inclusion of further terms". (which are not included in 132-137, I suppose.) 139 equal, Bubar=Bdbar, resulting in a single... 153 How did you assign the variations that were used? 177 etc fixed points? 182 Clarify or omit sentence. 204 Omit "It was checked that" 211 is differentd -> differs 215 Somehow this paragraph seems to be left hanging. 224 ... uncertainties", which.... 226 This source of uncertainty [suddenly "uncertainty" become singular.] 234 omit "so-called" but remind reader what this chisq means. 240 ... uncertainty, which dominates the uncertainties. 241 Eq. 7 doesn't seem to be there. 245 Omit the vague word "significantly" ? 249-252 Consider moving this paragraph up a bit, near to where the chisq is first discussed in this section. 256 Well there can be no doubt about the minima! I would simply say that the positions of minimum chisquared are in close agreement, indicating that any anomalies at low Q^2 are small. 258 This is hardly surprising since inclusive DIS is dominated by a QED vertex. It would be as well to remind the reader of this, perhaps at the start of the section where you correctly say that jets are essential. Does fig 2c add anything to the paper? 271 analyses 277 which is similar to the cut used here. 278 ... that both the H1 and...were performed 286 Consider just calling it "NNLO PDF evaluation" 297 I'd rearrange: it becomes questionable to quantify the theory uncertainties.... 302 Not sure what "map out" means. 333 entered -> used 337, 358 Omit very 340 For completeness add "DIS" into the sentence. 342 ... have been used... 343 called -> denoted as 345 Maybe: ... with alphaS allowed to vary freely.. 347 This result for... 350 for -> with Fig 7, 8. You might consider replacing the green by light blue to assist the colour-blind. (I add that this does not include me) However if you had no complaints before, perhaps this is not so important. Comment: A first reaction is why not include the PDF set with floating alphaS and uncertainties associated with this method. I can understand a rationale: a future user might want to use yet another value of alphaS, perhaps with an uncertainty, and hopefully will be able to extract the desired information from the two sets you have provided. It might be good to make this explicit. ______________________________________________________________________ Robert Klanner Comments to: “Impact of jet production data on the next-to-next-to-leading order5 determination of HERAPDF2.0 parton distributions” Paper Draft v0.5 – July 27, 2020 The paper is very well written, contains highly relevant results and is very complete. My comments refer to minor formal issues l.24 “agree very well” For me “very well” is a very subjective judgement; they either agree, possibly specified by “within statistical uncertainties” or “within systematic uncertainties”, or the do not agree. Similar, the “excellent” in l.98. I have no problems, if this comment is not taken into account. l.271 both analysis -> both analyses l.262 “+/-0.009(exp)” The spacing between the number and the opening parenthesis appears to be inconsistent in the paper. Personally I prefer a fixed small space. References [17] Is “H. Collaboration” correct? [29] F. Aaron et al., [H1 and ZEUS Collaboration], -> F. Aaron et al. [H1 and ZEUS Collaboration], -> remove “,” for consistency. Table 2 : Central Value -> Central value Fig. 8 : I would have expected that delta(xg)/xg is centered at 0 and not at 1. What am I missing? I probably I have not understood “normalised uncertainty”. Fig.11 The variable _2 denote -> The variable _2 denotes l. 438 NNLO. at the -> NNLO. At the l.439 as the NNLO fit since the H1 -> as the NNLO fit, since the H1 Footnote p.25: I also agree with the inclusion of the appendix; I understand that this footnote will be removed. l.469 Additional Material -> Additional material; Alpha Scan -> Alpha scan Footnote p.27: I support the inclusion of the material in the paper. Caption Fig. 16, 17: old procedure -> procedure of Ref. [] In several of the captions (e.g. Fig. 15) both “gluon PDFs” and “gluon distributions” are used. In my opinion there should be only one term. I favour “gluonPDF”. Again, my congratulations to the authors for this excellent paper. Greetings, Robert _______________________________________________________________________________ ____________________H1__________________________________________________________ Joerg Gayler Comments to v0.5 July 27 Thank you very much for these efforts for final words of the two collaborations. Significant: I list here all questions that are not just wording only. ============== General: I miss a bit a general outline of the fit procedure. We determine alpha_s and PDFs. In the introduction one learns that we determine both simultaneously. Then in 3.1 to 3.4 nothing is said on alpha_s. Then alpha_s is determined fitting also pdfs, but discussed only in context of scaling violations. And then alpha_s is unexpectedly fixed to determine pdfs which one had assumed to have been already determined together with alpha_s. To understand the paper more easily, I would like to have some guidance early on. Comparisons with other results in section 4.1 is a bit meager, showing only the level of consistency, if data are treated similarly. But the difference of the Results or the agreement is of interest in first place. The paper could also compare the result to other important analyses in different reactions. What is the relevance of the result on alpha_s? Comments in detail: 3.1 Choice .. ---------------- 127 "extra parameters .. one at a time": Is there some arbitrariness in which sequence the further parameters were chosen? Is the actual choice of parameters depending on the sequence? 3.2. Model .. ---------------- What is done with alpha_s in these PDF fits? Or the other way round, one wonders whether these variations lead only to uncertainties on the PDFs and not on alpha_s. In the introduction it is said, that fitted together. 153 what determined the variation of mu^2_f0? 3.3 ----- 175 Here in 3.3 alpha_s is mentioned. But what in 3.2? I already asked there. 191 but mu_f0 = sqrt(1.6) = 1.26 < 1.37. One could check and state what happens then. 4 ---- 228 I guess (2,0 mu_r, 1,0 mu_F) 4.1 ----- 269-277 The reader ist not so much interested whether data agree, if treated the same way, but at least as much whether the results are very different or not if different assumptions are used. So the main results as they were presented should be compared as well. 282 Again the main focus seems to be on trying to do the same. 5 ----- 305 if mentioned, what was then actually done, which scale variations assumed, as for the alpha_s fit? 318-320 "The reduction in model and parameterisation uncertainty ... mostly due to the necessity to change the estimation procedure." What does this mean? Understandable? Further discussion is rather technical. I would expect some words on the actual observations, like that the largest effect is seen at large x. Is it really significant as it appears to be? Minor: ========= Abstract ----------- 24 "Predictions": a bit much, as fitted to these data. Suggest: calculated cross sections Data ------- 73 "excluded for several reasons, including..."Sounds as if there would be more reasons than explained. May be "for several reasons: to ensure ..." 3.2 ----- 153 "was added to the parameterization uncertainty". But there is nothing yet to add on. The reader thinks to have missed something. 3.3 ---- 165 parameterS 178 As I understand, it is more clear to say: "one of the mass-parameter values. ... was used ... 182 Is it obvious, the the M_b plot demonstrates the power of the method? Is it, because we get a parabola? 241 where is equation 7? 4. --- 264 multiplied into the .. gluon term. Is this just a factor (1 + Dx)? Is this a clear wording at least for natives? 4.1 ---- 278-280 Reads strange: first it is said that H1 and NNLOJet analyses were done using fixed PDFs. Next sentence a simultaneous fit of alpha_s and PDFs. 5 ---- 321. IN section 3.3 References ------------ Ref. 2 Was there some agreement on the sequence: "ZEUS and H1"? Why not alphabetical? Ref 36 (2019) ---> (2017) All the best, Joerg _________________________________________________________________________ Karla Cantún Dear Editors, Thank you for sharing the draft. In general, I enjoyed reading it. I just have a couple of minor comments: Reference [2] is cited twice on lines 35 and 36, which looks repetitive. Equation 7 is referenced on line 241, but I could not find it in the text. [fixed] Cheers, Karla Cantún _________________________________________________________________________________ Comments to the draft v.05-July 27, 2020 Impact of jet production on the NNLO determination of HERAPDF2.0 parton distributions Dieter Haidt, August 24, 2020 The present analysis of the H1 and ZEUS data from HERA based on the inclusive processes as well as jet production within NNLO of pQCD is a major achievement and constitutes an essential part of HERA’s legacy. Some of the authors remember the very first steps performed by MARK at SLAC and Gargamelle at CERN, half a century ago, and appreciate the huge effort to arrive at the mature results of HERA. The huge progress on the experimental side is complemented by the equally huge progress on the theory side allowing now for a comparison at NNLO level. The emphasis of the present analysis is to study the impact of data on jet production when added to the previous analyses of the inclusive data alone. Apart from increasing the statistics the unique value of jet production comes from the direct dependence of the gluon (lines 217-231). I suggest therefore devote the first chapter to discussing the improvement in the pdf parameters and their correlations determining the phenomenological parton distribution functions. The crux of all analyses is the limited knowledge of the gluon distribution function obtained from the triangular phasespace in (x,Q2). The jetdata provide a genuine new insight. Within the same framework : compare the parameter fit using the inclusive data alone with the new fit in this paper of inclusive data + jetdata. Illustrate it with two figures and make it quantitative with tables of the bestvalues and their correlations in both cases. Comments on the improved gluon distribution are welcome. Thanks to Katarzyna I had a chance to study the correlation table of the combined data. I had anticipated strong correlations the gluon and the seaquark parameters and also between alpha_s and the gluon parameters, but to my surprise there are only rather moderate correlations with all pdf parameters. Perhaps the comparison with the correlation table using the inclusive data alone will elucidate my prejudice. Line 139 : The momentum sumrule is an integral in x from 0 to 1. Given the conditions of our analysis the application of the momentum sumrule is useless and only introduces a bias. The reason is that an integration is required over the full range in x for all Q2. The behavior of the pdf at low x is unknown. Due to the triangular shape of the phase space the constraint for increasing Q2 is getting more and more uncertain, since there are no measurements for low x, while the parametric x -dependence merely reflects the evolved information from the low Q2-region. Therefore, drop the momentum sumrule. The heavy flavor contribution is included through pQCD evolution from massless quarks. A remark would be in order, why we did not use our own direct measurements. Is it because of missing NNLO predictions ? It is of particular interest to illustrate the impact of the systematic sources on the value of alpha_s. A figure showing the variation of alpha_s with the scale parameters would be revealing, since the uncertainty of the bestfit is dominated by the scale uncertainties. The parametric dependence should be done for fixed pdf, otherwise the effect may be compensated partially by adjusting the (many) pdf parameters. The point will be whether the value of alpha_s is significantly shifted when varying the scale parameters. If it turns out that the central value of alpha_s is little dependent on the scale uncertainties, then we have a strong argument that our determination in the spacelike region differs from the one in the timelike region, which is also affected by scale uncertainties. The determination of the parton distribution functions and alpha_s are linked. The key to the understanding of our paper is largely related to the chisquare function and the way it is exploited. I suggest to devote a detailed presentation, technicalities may be deferred to an Appendice. a. How is the chi2-function defined ? chi2(p) = Term with matrix(yi,f,C) + Term with constraints yi = measurements xi = kinematic variables f = prediction depending upon kinematics and free parameters C = correlation matrix containing the uncertainties p = alpha_s + pdf-parameters + systematic parameters b. How are the systematic uncertainties implemented ? and in particular the splitting ? c. Is the fit performed simultaneously for all free parameters ? Or in terms of fixed values for alpha_s ? Is the “so-called” chisquare scan a sequence of ch2-fits with fixed values of alpha_s including all other parameters as free, also the scale parameters ? d. The chi2/dof is quoted as 1.2. Does this indicate an underestimation of some uncertainties, perhaps the scale uncertainties ? If so, would it be possible to repeat the fit with enlarged uncertainties to see the effect on the bestvalue of alpha_s ? e. Is the meaning of chi2 the same everywhere in the text ? f. Is the gluon distribution function dominating value and size of alpha_s among the pdfs ? How big are the shifts in alpha_s caused by the systematics in the pdfs and the scales ? Shifts are linear effects. It may be justified to average over some of them. g. The determination of alpha_s is dominated by the systematic theory error of +-0.0024 compared to the experimental error +-0.001. One may ask whether the treatment of the systematics has an important effect on the central fit value of alpha_s (see point f) Line 284 : why is the scale uncertainty almost equal to this analysis (and not bigger) ? Figure 2a : I don’t understand. Are all parameters except for alpha_s free ? also the scales ? Note my remarks above. It seems to me vital to understand the interplay between alpha_s and the systematics, before any conclusion can be reached about the central value of alpha_s. Figures 3-7 : Why is the green band for medium x broad for uv and not for dv ? The figures for different alpha_s do not mean very much, since the pdfs are part of the fit. Figure 8 : I don’t understand. What remains constant and what is part of the fit ? What do you conclude ? Line 344 : are we quoting the fit values and correlations for the two values of alpha_s assumed. Line 347 : A careful discussion should be given qualifying “compatible”. Note my remarks above. I suggest a much extended conclusion. I have - for my own sake - tried to highlight what we have achieved., since the present paper is perhaps the last word and therefore a legacy of HERA inviting some remarks on the overall increase in knowledge as well as some critical remarks on further improvements left to future studies. We may comment on the performance of the collider HERA, the experimental achievements of H1 and ZEUS with detailed publications and a final comparison with QCD (for the first time) at NNLO. The confrontation of experiment and theory provides a consistent picture. The basic question is to what extent we have tested QCD. Our concern is perturbative QCD, which reduces the question to stating in which (x,Q2)-region pQCD is valid. This implies to worry about two frontiers (a) the transition in Q2 from the nonperturbative to perturbative regime and (b) the transition from moderate x-values to very low x-values. The first question is addressed by considering various starting scales in Q2, thus getting safe against higher twists. The other question, though important, may remain disregarded, since it concerns only the small tip in the phase space region at low Q2 and low x. Given the assumption that in the selected phase space region pQCD be valid the data are used to determine the parton distribution function, which are the necessary input to predict the observables. Since pQCD is applied at order NNLO (which is an achievement in itself), there is an intrinsic purely theoretical uncertainty coming from the truncation of the perturbative series. We address this uncertainty by varying the scales involved. Our conclusion about the impact of these systematic uncertainties will be decisive in judging the difference of our determination in the spacelike regime of alpha_s and the existing one in the timelike regime. In any case here is a task for future work in theory. The determination of the parton distributions is of value in itself and represents an important achievement, although there are still several weak aspects : the a priori assumption of the shapes, specific assumptions regarding the flavours , in particular the unknown s sbar-quark contribution and the role of the gluon. It should be emphasized that all in all the present knowledge is remarkable. The experimental information is shared between the determination of the parton distribution functions and the confrontation with pQCD. Is is possible to make an educated guess what fraction of the experimental information is actually available for testing pQCD ? The observation that the correlations between the pdf parameters and alpha_s are weak, is perhaps a strong argument in favour of real test of pQCD. Our analysis demonstrates a consistent picture and complements the efforts in the timelike region. ______________________________________________________________________ Jan und Nelly Olson Dear members of the Editorial Board for the paper "Impact of Jet Production Data on the Next-to-Next-to-Leading Order Determination of HERAPDF2.0 Parton Distributions" We have now read through the draft 0.5 from 27.07.2020 of this paper and are quite happy, the paper is already close to completion in the current form. From our side there are no objections to the content. We understand that this is a paper for the specialists and that therefore the introduction and physics background presentation is rudimentary, with the body of the paper concentrating on the technical details and problems of the fitting procedure. The occasional use of technical jargon is also acceptable, when seen from this aspect. Nevertheless, a couple of remarks can be made: Line 35 describes this paper as the "completion" of the HERAPDF2.0 family. We think that this is too strong a word, since we see that there are possibilities for still further steps; explicitly we think of future NNLO calculations for heavy quark data involvement in the fitting. Thus, inclusive + jets + heavy quarks for the PDFNNLO fitting. The heavy quarks enter also as Vector Meson Production, accessible to QCD calculations. In this connection we like to mention the paper arXiv:1908.08398 by C.A.Flett et al., "How to include exclusive J/psi production data in global pdf analyses". We believe that inclusion of such HERA data will improve the uncertainty situation at low x for the gluon PDF, once the corresponding NLO/NNLO calculations are available. Thus, we would not too strongly stress that the current paper is final, as far as HERAPDFs go. Hopefully there is much more to come! Another comment: Chapter 2 describes in detail how several data points were exempted from the analysis, for various technical reasons. In the final results, one would like to see the predictions of the fitted PDFs for these points, which were NOT included in the fitting procedure. The paper makes the statement that there is good agreement between the fits and the fitted data, which naively is no surprise and only confirms that the inclusive and the jet data are compatible. However, the reader gets curious and wonders about the agreement (or possible disagreement) in the immediate neighborhood of the fitting phase space, the neighborhood which is given by the left out data points. It remains to be said that the paper is very well written, the English is excellent and there are almost no typos. Congratulations! Our detailed remarks follow below. Our best wishes for a speedy and smooth publication! Nelly and Jan There are 8 footnotes, which all (except footnotes 7,8) need a small correction: "word \footnote{...}" --> "word\footnote{...}" in order not to "fly free in the air". In footnote 3 the final period "." is missing (line 122) We understand that footnotes 7,8 will not appear in the final paper. line 33 "milestone for" --> "milestone in" line 43 "charm and bottom masses" --> "charm- and bottom-quark masses" (cf. line 105) line 74 "and that the NNLO..." --> "and limiting the NNLO scale uncertainties from becoming too large" line 75 "to the ~24%" --> "to ~24%" line 90 "for Q2 starting" --> "for Q2 values starting" line 244 "hope substantially to reduce" --> "hope to substantially reduce" line 253 "data with relatively low Q2" --> "data with relatively low Q2 values" OR "data at relatively low Q2" line 287 "fit where" --> "fit in which" line 297 "on them" "on PDFs" line 321 "discussed Section" --> "discussed in Section" line 334 "the the" --> "the" Ref.[12] The given arXiv number, 1801.06415, is wrong and a repetition from ref.[11]. The correct arXiv number is [hep-ph/0609285] Ref.[13] The arXiv number 1208.3641 refers to a writeup by D.Britzger, K.Rabbertz, F.Stober and M.Wobisch, from DIS2012 with title "New features in version 2 of the fastNLO project" We believe that this is the correct reference, since it goes together with Ref.[12]. Thus, you have got the authors wrong here! Ref.[29] "et al.," --> "et al." Refs [9,10,12,15,33] ", and" --> " and" ______________________________________________________________________ Stefan Schmidt Dear authors, dear editor, congratulations for finishing up this draft. Here are a few comments. All the best, Stefan Paper structure and changes to figures/analysis =============================================== Section 4 and 5/6 should be swapped. First we determine PDFs with fixed alpha_s and compare to jets data. The free-alpha_s fits should come after that (same order as in the summary). ** As already said in the paper presentation, I would prefer very much to have figure 8 or 15 with the same alpha_s for both PDFs. Using different alpha_s blurs the message of the figure. People will possibly try to relate the uncertainty differences to the alpha_s choice rather than focus on the message of uncertainty improvements through the new data. minor textual comments ====================== line 38: a typo "Yhis" -> "This" line 85: maybe explain what is meant by "complete NNLO" (something like this ... consistently including both massive and light flavoured jets...) line 116: missing reference ... improved ... fits, confirming earlier findings [36]. line 188: perhaps explain this problem a bit more in detail? ... such that a non-vanishing charm contribution can be generated perturbatively for any scale above the M_c threshold. -------------------------------------------------------------------------- ** Mark Sutton on comment by Stefan Hi, on the note of differing alpha_s for the plots, I would disagree strongly - for the main comparison, we should compare the deafult results from the earlier fit, with the default result from the new fit. If it is felt necessary to address this point about the differing alphas values from the fits in this way, one could include additional plots comparing the fits, but using the same alphas value, as auxilliary plots, or in an appendix. In this case, it would be open to discussion whether the alphas value used should be that from the older fit, or from the newer fit - one could easily argue, that the older fit should be presented with the newer alphas value from the new fit, and indeed, since the new fit in some sense supercedes the older fit, and so might be expected to be in some sense "better", then this might be the more reasonable suggestion. In either case, using a value that was not the result from the fit itself might be expected to artifically inflate the errors, which may be another reason to disfavour doing this, but in that case, probably better to artificially inflate the errors of the older fit. However, clearly showing the results of a fit which included some parameter, but with the PDF then evaluated with a different value for that parameter, which was not the value of the parameter that resulted from the fit would be reasonably meaningless, since that would not actually be the result of the fit at all. So if the fits themselves favour particular alphas values, then showing the results of one of the fits with respect to a different alphas value would be highly misleading. If one wanted a comparison with the same alphas value, then both fits would need to be explicitly performed with the same value of alphas from the outset, but this would be a completely different comparison from comparing uncertainties from fits where the alphas was included as fit parameter, even if alphas itself was then later chosen to be the same for the evaluation of the PDFs since the alphas and gluon are strongly correlated. Cheers Mark -------------------------------------------------------------------------- Mandy: Just a note of clarification to Stefan and Mark's comments. We do have new fits with both alpha-s=0.118 and 0.1155 both fixed alpha_s fits. The alpha_s =0.1155 is our preferred value from the free alphas fit but we are showing the uncertainties for the fixed alphas fits. We can provide these uncertainty comparisons for both values for the new fit and indeed Katarzyna has some plots ready to show next week. We cannot so easily go back and use the value alpha_s=0.1155 for the old fit as this means re-running fits, we had never used this precise value in the past. ______________________________________________________________________ Stephan during PP: wants the comparison of the gluon uncertainties for same alpha_s=0.118 Jets NNLO NNLO is in pap10/new_figures:newFigure8-ssComment.pdf Katarzyna I've made the plot that Stefan requested (the uncertainty comparison with the same alpha_s). Please have a look. Things do look similar to the present comparison, just slightly worse (as we've expected). ______________________________________________________________________ Zhiqing Zhang Dear editors and all, Thank for providing the paper draft. I have only one major comment. My understanding based on the discussions during the meetings is that the main focus of the paper is on the HERAPDF2.0 NNLO jet PDF set and the result of alpha_s is a byproduct and should be de-emphased. The current version of the paper draft does not reflect this. Indeed, if we would take this alpha_s result more seriously, one should provide/clarify its correlation with the previous results based on the similar data sets. Otherwise it won’t be easy for others when performing comparison or averaging. Cheers, Zhiqing ______________________________________________________________________ ratio Plots: https://www.desy.de/~kklimek/private/nnloJets/finalPlots/ratios/ _______________________________________________________________________