Parton distributions from high-precision collider data
NNPDF Collaboration, Richard D. Ball, Valerio Bertone, Stefano Carrazza, Luigi Del Debbio, Stefano Forte, Patrick Groth-Merrild, Alberto Guffanti, Nathan P. Hartland, Zahari Kassabov, José I. Latorre, Emanuele R. Nocera, Juan Rojo, Luca Rottoli, Emma Slade, Maria Ubiali
NNNPDF
CAVENDISH-HEP-17-06CERN-TH-2017-077Edinburgh 2017/08Nikhef/2017-006OUTP-17-04PTIF-UNIMI-2017-3
Parton distributions from high-precision collider data
The NNPDF Collaboration:
Richard D. Ball, Valerio Bertone, Stefano Carrazza, Luigi Del Debbio, Stefano Forte, Patrick Groth-Merrild, Alberto Guffanti, Nathan P. Hartland, Zahari Kassabov, , Jos´e I. Latorre, , Emanuele R. Nocera, Juan Rojo, Luca Rottoli, Emma Slade, and Maria Ubiali The Higgs Centre for Theoretical Physics, University of Edinburgh,JCMB, KB, Mayfield Rd, Edinburgh EH9 3JZ, Scotland Department of Physics and Astronomy, VU University, NL-1081 HV Amsterdam,and Nikhef Theory Group, Science Park 105, 1098 XG Amsterdam, The Netherlands Theoretical Physics Department, CERN, CH-1211 Geneva, Switzerland Tif Lab, Dipartimento di Fisica, Universit`a di Milano andINFN, Sezione di Milano, Via Celoria 16, I-20133 Milano, Italy Dipartimento di Fisica, Universit`a di Torino andINFN, Sezione di Torino, Via P. Giuria 1, I-10125, Turin, Italy Departament de F´ısica Qu`antica i Astrof´ısica, Universitat de Barcelona,Diagonal 645, 08028 Barcelona, Spain Center for Quantum Technologies, National University of Singapore Rudolf Peierls Centre for Theoretical Physics, 1 Keble Road,University of Oxford, OX1 3NP Oxford, United Kingdom Cavendish Laboratory, HEP group, University of Cambridge,J.J. Thomson Avenue, Cambridge CB3 0HE, United Kingdom a r X i v : . [ h e p - ph ] S e p bstract We present a new set of parton distributions, NNPDF3.1, which updates NNPDF3.0, the firstglobal set of PDFs determined using a methodology validated by a closure test. The updateis motivated by recent progress in methodology and available data, and involves both. Onthe methodological side, we now parametrize and determine the charm PDF alongside the lightquarks and gluon ones, thereby increasing from seven to eight the number of independent PDFs.On the data side, we now include the D0 electron and muon W asymmetries from the finalTevatron dataset, the complete LHCb measurements of W and Z production in the forwardregion at 7 and 8 TeV, and new ATLAS and CMS measurements of inclusive jet and electroweakboson production. We also include for the first time top-quark pair differential distributions andthe transverse momentum of the Z bosons from ATLAS and CMS. We investigate the impactof parametrizing charm and provide evidence that the accuracy and stability of the PDFs arethereby improved. We study the impact of the new data by producing a variety of determinationsbased on reduced datasets. We find that both improvements have a significant impact on thePDFs, with some substantial reductions in uncertainties, but with the new PDFs generally inagreement with the previous set at the one sigma level. The most significant changes are seen inthe light-quark flavor separation, and in increased precision in the determination of the gluon.We explore the implications of NNPDF3.1 for LHC phenomenology at Run II, compare withrecent LHC measurements at 13 TeV, provide updated predictions for Higgs production cross-sections and discuss the strangeness and charm content of the proton in light of our improveddataset and methodology. The NNPDF3.1 PDFs are delivered for the first time both as Hessiansets, and as optimized Monte Carlo sets with a compressed number of replicas.2 ontents Z bosons . . . . . . . . . . . . . . . . . . . . . . . . 152.7 Differential distributions and total cross-sections in t ¯ t production . . . . . . . . . 17 Z boson . . . . . . . . . . . . . . . . . . . . . . 424.3 Differential distributions for top pair production . . . . . . . . . . . . . . . . . . 434.4 Inclusive jet production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434.5 Electroweak boson production in the forward region . . . . . . . . . . . . . . . . 514.6 W asymmetries from the Tevatron . . . . . . . . . . . . . . . . . . . . . . . . . . 524.7 The ATLAS W, Z production data and strangeness . . . . . . . . . . . . . . . . . 534.8 The CMS 8 TeV double-differential Drell-Yan distributions . . . . . . . . . . . . 544.9 The EMC F c data and intrinsic charm. . . . . . . . . . . . . . . . . . . . . . . . 564.10 The impact of LHC data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594.11 Nuclear targets and nuclear corrections. . . . . . . . . . . . . . . . . . . . . . . . 604.12 Collider-only parton distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 W and Z production at the LHC 13 TeV . . . . . . . . . . . . . . . . . . . . . . 805.6 Higgs production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 A Code development and benchmarking 91 Introduction
A precise understanding of parton distributions [1–3] (PDFs) has played a major role in thediscovery of the Higgs boson and will be a key ingredient in searches for new physics at theLHC [4]. In recent years a new generation of PDF sets [5–11] have been developed for use at theLHC Run II. Some of these have been used in the construction of the PDF4LHC15 combined sets,recommended for new physics searches and for the assessment of PDF uncertainties on precisionobservables [12]. These PDF4LHC15 sets are obtained by means of statistical combination ofthe three global sets [5–7]: this is justified by the improved level of agreement in the globaldeterminations, with differences between them largely consistent with statistical fluctuation.Despite these developments, there remains a need for improvements in the precision andreliability of PDF determinations. Precision measurements at the LHC, such as in the searchfor new physics through Higgs coupling measurements, will eventually require a systematicknowledge of PDFs at the percent level in order to fully exploit the LHC’s potential. TheNNPDF3.0 PDF set [5], which is one of the sets entering the PDF4LHC15 combination, isunique in being a PDF set based on a methodology systematically validated by means of closuretests, which ensure the statistical consistency of the procedure used to extract the PDFs fromdata. The goal of this paper is to present NNPDF3.1, an update of the NNPDF3.0 set, and afirst step towards PDFs with percent-level uncertainties. Two directions of progress are requiredin order to reach this goal, the motivation for an update being accordingly twofold.On the one hand, bringing the precision of PDFs down to the percent level needs a largerand more precise dataset, with correspondingly precise theoretical predictions. In the time sincethe release of NNPDF3.0, a significant number of new experimental measurements have becomeavailable. From the Tevatron, we now have the final measurements of the W boson asymmetrieswith the electron and muon final states based upon the complete Run II dataset [13, 14]. At theLHC, the ATLAS, CMS and LHCb experiments have released a wide variety of measurementson inclusive jet production, gauge boson production and top production. Finally, the combinedlegacy measurements of DIS structure functions from HERA have also become available [9]. Inparallel with the experimental developments, an impressive number of new high-precision QCDcalculations of hadron collider processes with direct sensitivity to PDFs have recently beencompleted, enabling their use in the determination of PDFs at NNLO. These include differentialdistributions in top quark pair production [15, 16], the transverse momentum of the Z and W bosons [17, 18], and inclusive jet production [19, 20], for all of which precision ATLAS and CMSdatasets are available.All of these new datasets and calculations have been incorporated into NNPDF3.1. Theinclusion of the new data presents new challenges. Given the large datasets on which some ofthese measurements are based, uncorrelated experimental uncertainties are often at the permillelevel. Achieving a good fit then requires an unprecedented control of both correlated systematicsand of the numerical accuracy of theoretical predictions.On the other hand, with uncertainties at the percent level, accuracy issues related to theo-retical uncertainties hitherto not included in PDF determinations become relevant. Whereas thecomprehensive inclusion of theoretical uncertainties in PDF determination will require furtherstudy, we have recently argued that a significant source of theoretical bias arises from the conven-tional assumption that charm is generated entirely perturbatively from gluons and light quarks.A methodology which allows for the inclusion of a parametrized heavy quark PDFs within theFONLL matched general-mass variable flavor number scheme has been developed [21, 22], andimplemented in an NNPDF PDF determination [23]. It was found that when the charm PDF isparametrized and determined from the data alongside the other PDFs, much of the uncertaintyrelated to the value of the charm mass becomes part of the standard PDF uncertainty, while anybias related to the assumption that the charm PDF is purely perturbative is eliminated [23]. In4NPDF3.1 charm is therefore parametrized as an independent PDF, in an equivalent manner tolight quarks and the gluon. We will show that this leads to improvements in fit quality withoutan increase in uncertainty, and that it stabilises the dependence of PDFs on the charm mass,all but removing it in the light quark PDFs.The NNPDF3.1 PDf sets are released at LO, NLO, and NNLO accuracy. For the first time, allNLO and NNLO PDFs are delivered both as Hessian sets and as Monte Carlo replicas, exploitingrecent powerful methods for the construction of optimal Hessian representations of PDFs [24].Furthermore, and also for the first time, the default PDF sets are provided as compressed MonteCarlo sets [25]. Therefore despite being presented as sets of only 100 Monte Carlo replicas, theyexhibit many of the statistical properties of a much larger set, reducing observable computationtime without loss of information. A further improvement in computational efficiency can beobtained by means of the SM-PDF tool [26], which allows for the selection of optimal subsetsof Hessian eigenvectors for the computation of uncertainties on specific processes or classes ofprocesses, and which is available as a web interface [27] now also including the NNPDF3.1sets. A variety of PDF sets based on subsets of data are also provided (as standard 100 replicaMonte Carlo sets), which may be useful for specific applications such as new physics searches,or measurements of standard model parameters.The outline of this paper is as follows. First, in Sect. 2 we discuss the experimental aspectsand the relevant theoretical issues of the new datasets. We then turn in Sect. 3 to a detaileddescription of the baseline NNPDF3.1 PDF sets, with a specific discussion of the impact ofmethodological improvements, specifically the fact that the charm PDF is now independentlyparametrized and determined like all other PDFs. In Sect. 4 we discuss the impact of the newdata by comparing PDF sets based upon various data subsets, and also discuss PDF sets basedon more conservative data subsets. In Sect. 5 we summarise the status of uncertainties on PDFsand luminosities, and specifically discuss the strange and charm content of the proton in light ofour results, and present first phenomenological studies at the LHC. Finally, a summary of thePDFs being delivered in various formats is provided in Sect. 6, together with links to repositorieswhence more detailed sets of plots may be downloaded.
The NNPDF3.1 PDF sets include a wealth of new experimental data. We have augmented ourdataset with improved determinations of observables already included in NNPDF 3.0 (such as W and Z rapidity distributions) as well as two new process: top quark differential distributions,and the Z transverse momentum distribution, which is include for the first time in a global PDFdetermination.In this Section we discuss the NNPDF3.1 dataset in detail. After a general overview, eachobservable will be examined: we describe the individual measurements, and address specifictheoretical and phenomenological issues related to their inclusion, particularly in relation to theuse of recent NNLO results.In NNPDF3.1 only LHC data from Run I, taken at centre-of-mass energies of 2.76 TeV,7 TeV and 8 TeV (with one single exception), are included. The more recent 13 TeV datasetis reserved for phenomenological comparison purposes in Sect. 5. Available and upcoming LHCRun II data at 13 TeV will be part of future NNPDF releases. The NNPDF3.0 global analysis involved data from deep-inelastic scattering (DIS) experiments,fixed-target Drell-Yan data, and collider measurements from the Tevatron and LHC. The fixed-target and collider DIS datasets included measurements from NMC [28,29], BCDMS [30,31] and5LAC [32]; the combined HERA-I inclusive structure function dataset [33] and HERA-II inclu-sive measurements from H1 and ZEUS [34–37]; the HERA combined measurements of the charmproduction cross-section σ NC c [38]; CHORUS inclusive neutrino DIS [39], and NuTeV dimuonproduction data [40, 41]. From the Tevatron, CDF [42] and D0 [43] Z rapidity distributions;and CDF [44] Run-II one-jet inclusive cross-sections were used. Constraints from fixed-targetDrell-Yan came from the E605 [45] and E866 [46–48] experiments. LHC measurements includedelectroweak boson production data from ATLAS [49–51], CMS [52–54] and LHCb [55, 56]; one-jet inclusive cross-sections from ATLAS [57, 58] and CMS [59]; the differential distributions for W production in association with charm quarks from CMS [60]; and total cross-section mea-surements for top quark pair production data from ATLAS and CMS at 7 and 8 TeV [61–66].For NNPDF3.1 we have made a number of improvements to the NNPDF3.0 dataset. Firstlywe have included the final datasets for several experiments which have now concluded, replacingsuperseded data in the NNPDF3.0 analysis. The HERA-I data and the H1 and ZEUS HERA-IIinclusive structure functions have been replaced by the final HERA combination [9]. The HERAdataset has also been enlarged by the inclusion of H1 and ZEUS measurements of the bottomstructure function F b ( x, Q ) [67,68], which may prove useful in specific applications such as in thedetermination of the bottom quark mass m b . In order to perform dedicated studies of the charmcontent of the proton, we have constructed a PDF set also including the EMC measurements ofcharm structure functions at large- x [69], which will be discussed in Sect. 5.3. However, thesemeasurements are not included in the standard dataset. The legacy W lepton asymmetriesfrom D0 using the complete Tevatron luminosity, both in the electron [14] and in the muon [13]channels have been added. These precise weak gauge boson production measurements provideimportant information on the quark flavor separation at large- x , as demonstrated in [70].Aside from the updated legacy datsets, in NNPDF3.1 a large number of recent measurementsfrom ATLAS, CMS and LHCb are included. For ATLAS, we now include the Z boson ( p ZT , y Z )and ( p ZT , M ll ) double differential distributions measured at 8 TeV [71]; the inclusive W + , W − and Z rapidity distributions at 7 TeV from the 2011 dataset [72], the top-quark pair productionnormalized y t distribution at 8 TeV [73]; total cross-sections for top quark pair production at 7,8 and 13 TeV [74,75]; inclusive jet cross-sections at 7 TeV from the 2011 dataset [76]; and finallylow mass Drell-Yan M ll distributions at 7 TeV from the 2010 run [77]. The transverse momentumspectrum at 7 TeV (2011 dataset) [78] will be studied in Sec. 4.2 but it is not included in thedefault set. The total top cross-section is the only data point at 13 TeV which is included. ForCMS, NNPDF3.1 includes the W + and W − rapidity distributions at 8 TeV [79], together withtheir cross-correlations; the inclusive jet production cross-sections at 2.76 TeV [80]; top-quarkpair production normalized y t ¯ t distributions at 8 TeV [81], total inclusive t ¯ t cross-sections at 7,8 and 13 TeV [82]; the distribution of the Z boson double differentially in ( p T , y Z ) at 8 TeV [83].The double-differential distributions ( y ll , M ll ) in Drell-Yan production at 8 TeV [84] will bestudied in Sect. 4.8 below, but it is not included in the default PDF determination. For LHCb,NNPDF3.1 includes the complete 7 and 8 TeV measurements of inclusive W and Z productionin the muon channel [85,86], which supersedes all previous measurements in the same final state.An overview of the data included in NNPDF3.1 is presented in Tables 2.1, 2.2, and 2.3, forthe DIS structure function data, the fixed target and Tevatron Drell-Yan experiments, and theLHC datasets, respectively. For each dataset we indicate the corresponding published refer-ence, the number of data points in the NLO/NNLO PDF determinations before and after (inparenthesis) kinematic cuts, the kinematic range covered in the relevant variables after cuts,and the code used to compute the NLO and NNLO results. Datasets included for the first timein NNPDF3.1 are flagged with an asterisk. The datasets not used for the default determina-tion are in brackets. The total number of data points for the default PDF determination is4175 / / (cid:0) x, Q (cid:1) plane.6 xperiment Obs. Ref. N dat x range Q range (GeV) TheoryNMC F d /F p [28] 260 (121/121) 0 . ≤ x ≤ .
68 2 . ≤ Q ≤ APFEL σ NC , p [29] 292 (204/204) 0 . ≤ x ≤ .
50 1 . ≤ Q ≤ . F p [32] 211 (33/33) 0 . ≤ x ≤ .
55 1 . ≤ Q ≤ . APFEL F d [32] 211 (34/34) 0 . ≤ x ≤ .
55 1 . ≤ Q ≤ . F p [30] 351 (333/333) 0 . ≤ x ≤ .
75 2 . ≤ Q ≤ . APFEL F d [31] 254 (248/248) 0 . ≤ x ≤ .
75 3 . ≤ Q ≤ . σ CC ,ν [39] 607 (416/416) 0 . ≤ x ≤ .
65 1 . ≤ Q ≤ . APFEL σ CC , ¯ ν [39] 607 (416/416) 0 . ≤ x ≤ .
65 1 . ≤ Q ≤ . σ ccν [40, 41] 45 (39/39) 0 . ≤ x ≤ .
33 2 . ≤ Q ≤ . APFEL σ cc ¯ ν [40, 41] 45 (37/37) 0 . ≤ x ≤ .
21 1 . ≤ Q ≤ . σ p NC , CC (*) [9] 1306 (1145/1145) 4 · − ≤ x ≤ .
65 1 . ≤ Q ≤ APFEL σ c NC [38] 52 (47/37) 7 · − ≤ x ≤ .
05 2 . ≤ Q ≤ F b (*) [67, 68] 29 (29/29) 2 · − ≤ x ≤ . . ≤ Q ≤ F c ] (*) [69] 21 (16/16) 0 . ≤ x ≤ .
44 2 . ≤ Q ≤ . APFEL
Table 2.1:
Deep-inelastic scattering data included in NNPDF3.1. The EMC F c data are in bracketsbecause they are only included in a dedicated set but not in the default dataset. New datasets, notincluded in NNPDF3.0, are denoted (*) . The kinematic range covered in each variable is given aftercuts are applied. The total number of DIS data points after cuts is 3102 / F c data). Exp. Obs. Ref. N dat Kin Kin (GeV) TheoryE866 σ d DY /σ p DY [48] 15 (15/15) 0.07 ≤ y ll ≤ .
53 4 . ≤ M ll ≤ . APFEL+Vrap σ p DY [46, 47] 184 (89/89) 0 ≤ y ll ≤ .
36 4 . ≤ M ll ≤ . APFEL+Vrap
E605 σ p DY [45] 119 (85/85) − . ≤ y ll ≤ . . ≤ M ll ≤ . APFEL+Vrap
CDF dσ Z /dy Z [42] 29 (29/29) 0 ≤ y ll ≤ . ≤ M ll ≤ Sherpa+Vrap k t incl jets [87] 76 (76/76) 0 ≤ y jet ≤ . ≤ p jet T ≤ NLOjet++ D0 dσ Z /dy Z [43] 28 (28/28) 0 ≤ y ll ≤ . ≤ M ll ≤ Sherpa+Vrap W electron asy (*) [14] 13 (13/8) 0 ≤ y e ≤ . Q = M W MCFM + FEWZ W muon asy (*) [13] 10 (10/9) 0 ≤ y µ ≤ . Q = M W MCFM + FEWZ
Table 2.2:
Same as Table 2.1 for the Tevatron fixed-target Drell-Yan and W , Z and jet collider data.The total number of Tevatron data points after cuts is 345 /
339 for NLO/NNLO fits. xp. Obs. Ref. N dat Kin Kin (GeV) TheoryATLAS W, Z ≤ | η l | ≤ . Q = M W , M Z MCFM + FEWZ
W, Z (*) [72] 34 (34/34) 0 ≤ | η l | ≤ . Q = M W , M Z MCFM + FEWZ high-mass DY 2011 [50] 11 (5/5) 0 ≤ | η l | ≤ . ≤ M ll ≤ MCFM + FEWZ low-mass DY 2011 (*) [77] 6 (4/6) 0 ≤ | η l | ≤ . ≤ M ll ≤ MCFM + FEWZ [ Z p T (cid:0) p ZT , y Z (cid:1) ] (*) [78] 64 (39/39) 0 ≤ | y Z | ≤ . ≤ p ZT ≤ MCFM +NNLO
Z p T (cid:0) p ZT , M ll (cid:1) (*) [71] 64 (44/44) 12 ≤ M ll ≤
150 GeV 30 ≤ p ZT ≤ MCFM +NNLO
Z p T (cid:0) p ZT , y Z (cid:1) (*) [71] 120 (48/48) 0 . ≤ | y Z | ≤ . ≤ p ZT ≤ MCFM +NNLO7 TeV jets 2010 [57] 90 (90/90) 0 ≤ | y jet | ≤ . ≤ p jet T ≤ NLOjet++ ≤ | y jet | ≤ . ≤ p jet T ≤ NLOjet++ (*) [76] 140 (31/31) 0 ≤ | y jet | ≤ . ≤ p jet T ≤ NLOjet++ σ tot ( t ¯ t ) [74, 75] 3 (3/3) - Q = m t top++ (1 /σ t ¯ t ) dσ ( t ¯ t ) /y t (*) [73] 10 (10/10) 0 < | y t | < . Q = m t Sherpa +NNLOCMS W electron asy [52] 11 (11/11) 0 ≤ | η e | ≤ . Q = M W MCFM + FEWZ W muon asy [53] 11 (11/11) 0 ≤ | η µ | ≤ . Q = M W MCFM + FEWZ W + c total [60] 5 (5/0) 0 ≤ | η l | ≤ . Q = M W MCFM W + c ratio [60] 5 (5/0) 0 ≤ | η l | ≤ . Q = M W MCFM
2D DY 2011 7 TeV [54] 124 (88/110) 0 ≤ | η ll | ≤ . ≤ M ll ≤ MCFM + FEWZ [2D DY 2012 8 TeV] [84] 124 (108/108) 0 ≤ | η ll | ≤ . ≤ M ll ≤ MCFM + FEWZ W ± rap 8 TeV (*) [79] 22 (22/22) 0 ≤ | η l | ≤ . Q = M W MCFM + FEWZ
Z p T (*) [83] 50 (28/28) 0 . ≤ | y Z | ≤ . ≤ p ZT ≤ MCFM +NNLO7 TeV jets 2011 [59] 133 (133/133) 0 ≤ | y jet | ≤ . ≤ p jet T ≤ NLOjet++ (*) [80] 81 (81/81) 0 ≤ | y jet | ≤ . ≤ p jet T ≤ NLOjet++ σ tot ( t ¯ t ) [82, 88] 3 (3/3) - Q = m t top++ (1 /σ t ¯ t ) dσ ( t ¯ t ) /y t ¯ t (*) [81] 10 (10/10) − . < y t ¯ t < . Q = m t Sherpa +NNLOLHCb Z rapidity 940 pb [55] 9 (9/9) 2 . ≤ η l ≤ . Q = M Z MCFM + FEWZ Z → ee rapidity 2 fb [56] 17 (17/17) 2 . ≤ η l ≤ . Q = M Z MCFM + FEWZ
W, Z → µ (*) [85] 33 (33/29) 2 . ≤ η l ≤ . Q = M W , M Z MCFM + FEWZ
W, Z → µ (*) [86] 34 (34/30) 2 . ≤ η l ≤ . Q = M W , M Z MCFM + FEWZ
Table 2.3:
Same as Table 2.1, for ATLAS, CMS and LHCb data from the LHC Run I at √ s = 2 .
76 TeV, √ s = 7 TeV and √ s = 8 TeV. The ATLAS 7 TeV Z p T and CMS 2D DY 2012 are in brackets becausethey are only included in a dedicated study but not in the default PDF set. The total number of LHCdata points after cuts is 848 /
854 for NLO/NNLO fits (not including ATLAS 7 TeV
Z p T and CMS 2DDY 2012). For hadronic data, leading-order kinematics have been assumed for illustrative purposes, withcentral rapidity used when rapidity is integrated over and the plotted value of Q set equal tothe factorization scale. It is clear that the new data added in NNPDF3.1 are distributed ina wide range of scales and x , considerably extending the kinematic reach and coverage of thedataset.In Table 2.4 we present a summary of the kinematic cuts applied to the various processesincluded in NNPDF3.1 at NLO and NNLO. These cuts ensure that only data where theoreticalcalculations are reliable are included. Specifically, we always remove from the NLO datasetpoints for which the NNLO corrections exceed the statistical uncertainty. The further cutscollected in Table 2.4, specific to individual datasets, will be described when discussing eachdataset in turn. All computations are performed up to NNLO in QCD, not including electroweakcorrections. We have checked that with the cuts described in Table 2.4, electroweak correctionsnever exceed experimental uncertainties.The codes used to perform NLO computations will be discussed in each subsection below.8 ataset NLO NNLODIS structure functions W ≥ . W ≥ . Q ≥ . Q ≥ . HERA σ NC c (in addition) - Q ≥ (fitted charm)ATLAS 7 TeV inclusive jets 2011 | y jet | ≤ . | y jet | ≤ . τ ≤ . τ ≤ . | y/y max | ≤ . | y/y max | ≤ . W → lν asymmetries - | A l | ≥ . ≤ M ll ≤
200 GeV M ll ≤
200 GeV | y Z | ≤ . | y Z | ≤ . M ll ≥
30 GeV M ll ≥
30 GeVLHCb 7 TeV and 8 TeV
W, Z → µ - | y l | ≥ . Z p T ≤ p ZT ≤
500 GeV 30 GeV ≤ p ZT ≤
500 GeVATLAS
Z p T p T , M ll ) p ZT ≥
30 GeV p ZT ≥
30 GeVATLAS
Z p T p T , y Z ) 30 GeV ≤ p ZT ≤
150 GeV 30 GeV ≤ p ZT ≤
150 GeVCMS
Z p T p T , y Z ) 30 GeV ≤ p ZT ≤
170 GeV 30 GeV ≤ p ZT ≤
170 GeV | y Z | ≤ . | y Z | ≤ . Table 2.4:
Full set of kinematical cuts applied to the processes used for NNPDF3.1 PDF determinationat NLO and at NNLO. Only data satisfying the constraints in the table are retained. The experimentsin brackets are not part of the global dataset and only used for dedicated studies. The cut on the HERAcharm structure function data at NNLO is applied only when charm is fitted, and it is applied in additionto the other DIS kinematical cuts.
With the exception of deep-inelastic scattering, NNLO corrections are implemented by comput-ing at the hadron level the bin-by-bin ratio of the NNLO to NLO prediction with a pre-definedPDF set, and applying the correction to the NLO computation (see Sect. 2.3 of Ref. [5]). Forall new data included in NNPDF3.1, the PDF set used for the computation of these correctionfactors (often refereed to as K -factors, and in Ref. [5] as C -factors) is NNPDF3.0, except forthe CMS W rap 8 TeV and ATLAS W/Z 2011 entries of Tab. 2.3 for which published xFitter results have been used and the CMS 2D DY 2012 data for which MMHT PDFs have beenused [89] (see Sect. 2.5 below); the PDF dependence of the correction factors is much smallerthan all other relevant uncertainties as we will demonstrate explicitly in Sect. 2.7 below.9 x10 Q ( G e V ) Kinematic coverage
Fixed target DISCollider DISFixed target Drell-YanCollider Inclusive Jet ProductionCollider Drell-YanZ transverse momentumTop-quark pair productionBlack edge: New in NNPDF3.1
Figure 2.1:
The kinematic coverage of the NNPDF3.1 dataset in the (cid:0) x, Q (cid:1) plane. .2 Deep-inelastic structure functions The main difference between the NNPDF 3.0 and 3.1 DIS structure function datasets is thereplacement of the separate HERA-I and ZEUS/H1 HERA-II inclusive structure function mea-surements by the final legacy HERA combination [9]. The impact of the HERA-II data on aglobal fit which includes HERA-I data is known [5, 90–92] to be moderate to begin with; thefurther impact of replacing the separate HERA-I and HERA-II data used in NNPDF3.0 withtheir combination has been studied in [93] and found to be completely negligible.Additionally, the NNPDF3.1 dataset includes the H1 and ZEUS measurements of the bottomstructure function F b ( x, Q ) [67, 68]. While the F b dataset is known to have a very limited pull,the inclusion of this dataset is useful for applications, such as the determination of the bottommass [94].While it is not included in the default NNPDF3.1 dataset, the EMC data on charm structurefunctions [69] will also be used for specific studies of the charm content of the proton in Sect. 5.3.As discussed in Refs. [23, 95], the EMC dataset has been corrected by updating the BR( D → µ )branching ratio: the value used in the original analysis [69] is replaced with the latest PDGvalue [96]. A conservative uncertainty on this branching ratio of ±
15% is also included.The cuts applied to DIS data are as follows. As in NNPDF3.0, for all structure functiondatasets we exclude data with Q < . and W < . , i.e. the region where highertwist corrections might become relevant and the perturbative expansion may become unreliable.At NNLO we also remove F c data with Q < in order to minimize the possible impactof unknown NNLO terms related to initial-state charm (see below).The computation of structure functions has changed in comparison to previous NNPDFreleases. Indeed, in NNPDF3.0 the solution of the DGLAP evolution equations and the structurefunctions were computed with the internal NNPDF code FKgenerator [97, 98], based on theMellin space formalism. In NNPDF3.1, as was already the case in the charm study of Ref. [23],PDF evolution and DIS structure functions are computed using the
APFEL public code [99],based instead on the x -space formalism. The two codes have been extensively benchmarkedagainst each other, see App. A. DIS structure functions are computed at NLO in the FONLL-Bgeneral-mass variable flavor number scheme, and at NNLO in the FONLL-C scheme [100]. Allcomputations include target mass corrections.In NNPDF3.1 we now parametrize charm independently, and thus the FONLL GM-VFNhas been extended in order to include initial-state heavy quarks. This is accomplished using theformalism of Refs. [21, 22]. Within this formalism, a massive correction to the charm-initiatedcontribution is included alongside the contribution of fitted charm as a non-vanishing boundarycondition to PDF evolution. At NNLO this correction requires knowledge of massive charm-initiated contributions to the DIS coefficient functions up to O (cid:0) α S (cid:1) , which are currently onlyknown to O ( α S ) [101]. Therefore, in the NNLO PDF determination, the NLO expression forthis correction is used: this corresponds to setting the unknown O (cid:0) α S (cid:1) contribution to themassive charm-initiated term to zero. Such an approximation was used Ref. [23], where it wasshown that it is justified by the fact that even setting to zero the full correction (i.e. using theLO expression for the massive correction) has an effect which at the PDF level is much smallerthan PDF uncertainties (see in particular Fig. 10 of Ref. [23]).Finally, as in previous NNPDF studies, no nuclear corrections are applied to the deuteronstructure function and neutrino charged-current cross-section data taken on heavy nuclei, inparticular NuTeV and CHORUS. We will return to this issue in Sect. 4.11. In NNPDF3.1 we have included the same fixed-target Drell-Yan (DY) data as in NNPDF3.0,namely the Fermilab E605 and E866 datasets; in the latter case both the proton-proton data11nd the ratio of cross-sections between deuteron and proton targets, σ d DY /σ p DY are included.However, the kinematic cuts applied to these two experiments differ from those in NNPDF3.0,based on the study of [102], which showed that theoretical predictions for data points too close tothe production threshold become unstable. Requiring reliability of the fixed-order perturbativeapproximation leads to the cuts τ ≤ .
08 and | y/y max | ≤ . , (2.1)where τ = M ll /s and y max = − ln τ , with M ll the dilepton invariant mass distribution and √ s the center of mass energy of the collision.As in the case of DIS, NLO fixed-target Drell-Yan cross-sections were computed in NNPDF3.0using the Mellin-space FKgenerator code, while in NNPDF3.1 they are obtained using
APFEL .The two computations are benchmarked in App. A. NNLO corrections are determined using
Vrap [103]. Once more, as in previous NNPDF studies, no nuclear corrections are applied; againwe will return to this issue in Sect. 4.11 below.
Four single-inclusive jet cross-section measurements were part of the NNPDF3.0 dataset: CDFRun II k T [44], CMS 2011 [59], ATLAS 7 TeV 2010 and ATLAS 2.76 TeV, including correlationsto the 7 TeV data [57, 58]. On top of these, in NNPDF3.1 we also include the ATLAS 7 TeV2011 [76] and CMS 2.76 TeV [80] data. Some of these measurements are available for differentvalues of the jet R parameter; the values used in NNPDF3.1 are listed in Table 2.5. Dataset Ref. Jet RadiusCDF Run II k t incl jets [87] R = 0 . R = 0 . R = 0 . R = 0 . R = 0 . R = 0 . Table 2.5:
Values of the jet R parameter used for the jet production datasets included in NNPDF3.1. No cuts are applied to any of jet datasets included in NNPDF3.1, except for the ATLAS2011 7 TeV data, for which achieving a good description turns out to be impossible if all fiverapidity bins are included simultaneously. We can obtain a good agreement between data andtheory when using only the central rapidity bin, | η jet | < .
4. The origin of this state of affairsis not understood: we have verified that a reasonable description can be obtained if some ofthe systematic uncertainties are decorrelated, but we have no justification for such a procedure.We have therefore chosen to only include in NNPDF3.1 data from the central rapidity bin, | η jet | < . NLOjet++ [105] interfaced to
APPLgrid [106]. The jet p T is used as the central factorization and renormalization scale in allcases, as this choice exhibits improved perturbative convergence compared to other scale choicessuch as the leading jet p T [107, 108].While the NNLO calculation of inclusive jet production has been recently published [20,108],results are not yet available for all datasets included in NNPDF3.1. Therefore, jet data areincluded as default in the NNPDF3.1 NNLO determination using NNLO PDF evolution but NLOmatrix elements, while adding to the covariance matrix an additional fully correlated theoretical12 (GeV) T p
200 400 600 800 1000 1200 1400 1600 F r a c t i ona l N L O sc a l e un c e r t a i n t y ATLAS 2011 7 TeV incl jets, 0 < |y| < 0.5ATLAS 2011 7 TeV incl jets, 0 < |y| < 0.5 (GeV) T p
100 150 200 250 300 350 400 450 500 550 F r a c t i ona l N L O sc a l e un c e r t a i n t y CMS 2.76 TeV TeV incl jets, 0 < |y| < 0.5CMS 2.76 TeV TeV incl jets, 0 < |y| < 0.5
Figure 2.2:
The fractional scale uncertainty on NLO single-inclusive jet production, as a function of thejet p T for the central rapidity bins of ATLAS 7 TeV 2011 (left) and the CMS 2.76 TeV (right). systematic uncertainty estimated from scale variation of the NLO calculation. The NLO scalevariations are performed using APPLgrid interfaced to
HOPPET [109]. We take the associateduncertainty as the the envelope of the result of seven-point scale variation µ F ∈ [ p T / , p T ] and µ R ∈ [ p T / , p T ] with 1 / ≤ µ F /µ R ≤ p T is chosen as a central scale [108]. This scale uncertainty isshown in Fig. 2.2 for ATLAS 7 TeV 2011 and CMS 2.76 TeV as a function of the jet p T for thecentral rapidity bin. It is seen to range between a few percent at low p T up to around 10% atthe largest p T . A similar behaviour is observed in other rapidity bins, with a more asymmetricband at forward rapidity.In order to gauge the reliability of our approximate treatment of the jet data, we haveproduced a PDF determination in which all data for which NNLO corrections are known, namelythe 7 TeV ATLAS and CMS datasets, are included using exact NNLO theory. This will bediscussed in Sect. 4.4. Representative NNLO corrections are shown in Fig. 2.3, where we showthe NNLO/NLO ratio for the central rapidity bin (0 ≤ | y jet | ≤ .
5) of the ATLAS and CMS7 TeV 2011 datasets, plotted as a function of p T [110]: note (see Table 2.5) that the values of R are different, thereby explaining the different size of the correction, which for CMS is ∼ −
2% for p T ∼
100 GeV, increasing up to ∼
5% for p T ∼ ∼ − ∼
9% as a function of p T . Unlike in the case of the Z transverse momentumdistribution, to be discussed in Sect. 2.6, the lack of smoothness of the corrections seen in Fig. 2.3is not problematic as the fluctuations are rather smaller than typical uncorrelated uncertaintieson these data. The NNPDF3.0 determination already included a wide set of collider Drell-Yan data, both atthe W and Z peak and off-shell. This dataset has been further expanded in NNPDF3.1. Wediscuss here invariant mass and rapidity distributions; transverse momentum distributions willbe discussed in Sect. 2.6.In NNPDF3.1 we include for the first time D0 legacy W asymmetry measurements basedon the complete dataset in the electron [14] and muon [13] channels. The only cut applied tothis dataset is at NNLO, where we remove data with A l ( y l ) ≤ .
03 in both the electron andmuon channel data. This is due to the fact that when the asymmetry is very close to zero,even with high absolute accuracy on the NNLO theoretical calculation, it is difficult to achievehigh percentage accuracy, thereby making the NNLO correction to the asymmetry unreliable.The NLO computation is performed using
APPLgrids from the HERAfitter study of [70], whichwe have cross-checked using
Sherpa [111] interfaced to
MCgrid [112]. NNLO corrections are13 (GeV) T p NN L O / N L O K - f a c t o r s NNLO/NLO K-factors for inclusive jet production
CMS 7 TeV 2011 (R=0.7)ATLAS 7 TeV 2011 (R=0.6)
NNLO/NLO K-factors for inclusive jet production
Figure 2.3:
The NNLO/NLO cross-section ratio [110] for the central rapidity bin (0 ≤ | y jet | ≤ .
5) ofthe ATLAS and CMS 7 TeV 2011 jet data, with the values of R of Tab. 2.5, plotted vs. p T . computed using FEWZ [113–115].New results are included for ATLAS, CMS and LHCb. For ATLAS, NNPDF3.0 included 2010 W and Z − is includedin NNPDF3.1, albeit partially. This measurement provides differential distributions in leptonpseudo-rapidity | η l | in the range 0 ≤ | η l | ≤ . W + and W − production. For Z/γ ∗ production results are provided either with both leptons measured in the range 0 ≤ | η l | ≤ . ≤ | η l | ≤ . . ≤ | η l | ≤ .
9. The central rapiditydata sre given for three bins in the dilepton invariant mass 46 < m ll <
66, 66 < m ll <
116 and116 < m ll <
150 GeV, and the forward rapidity data in the last two mass bins (on-peak andhigh-mass). We only include the on-shell, 0 ≤ | η l | ≤ . Z production bins in the central rapidity region, and the on-peak and high-mass Z production bins at forward rapidity. The full dataset will be included in future NNPDFreleases. No other cuts are applied to the dataset. Theoretical predictions are obtained usingNLO APPLgrids [106] generated using
MCFM [116], while the NNLO corrections are taken fromthe xFitter analysis of Ref [72].Also new to NNPDF3.1 is the ATLAS low-mass Drell-Yan data from Ref. [77]. We use onlythe low-mass DY cross-sections in the muon channel measured from 35 pb − M ll = 12 GeV. The 2011 7 TeV data with invariant masses between 26 GeVand 66 GeV are not included because they are affected by large electroweak corrections andare therefore excluded by our cuts. Furthermore, two datapoints are removed from the NLOdatasets because NNLO corrections exceed experimental uncertainties. Theoretical predictionsare obtained at NLO using APPLgrids [106] constructed using
MCFM , and at NNLO correctionsare computed using
FEWZ .For CMS, NNPDF3.1 includes 8 TeV W + and W − rapidity distributions, including informa-tion on their correlation [79]. No cuts have been applied to this dataset. Theoretical predictionsare obtained using the NLO APPLgrids generated with
MCFM and the NNLO correction factorscomputed using
FEWZ in the context of the xFitter [117] analysis presented in Ref. [79]. Dou-ble differential rapidity y ll and invariant mass M ll distributions for Z/γ ∗ production from the14igure 2.4: The NNLO/NLO cross-section for the LHCb 7 (left) and 8 TeV (right) data. The centralrapidity region which is cut is shaded in red. M (cid:96)(cid:96) ≥
30 GeV, because in the lowest mass bin the leading-order prediction in this bin vanishes.Theoretical predictions are obtained at NLO using
APPLgrids constructed using
MCFM , and atNNLO corrections have been computed [89] using
FEWZ .For LHCb, previous data included in NNPDF3.0 are replaced by the final 7 TeV and8 TeV W +, W − and Z rapidity distributions in the muon channel [85, 86]. The NNLO/NLOcross-section ratios are shown in Fig. 2.4. The Data with | y l | ≤ .
25 from this set have been cutbecause the anomalously large size of the NNLO corrections suggests that they may be unreli-able. Theoretical predictions are obtained at NLO using
APPLgrids constructed using MCFM,and at NNLO corrections computed using
FEWZ . Z bosons The transverse momentum distribution of the Z boson is included for the first time in a globalPDF determination thanks to the recent computation of the process at NNLO [18, 118–120]. Inthe NNPDF3.1 determination we include recent datasets from ATLAS and CMS following thedetailed study in Ref. [121].ATLAS has published measurements of the spectrum of the Z transverse momentum at7 TeV [78] and at 8 TeV [71]. Measurements are performed in the Z/γ ∗ → e + e − and Z/γ ∗ → µ + µ − channels which are then combined. The 7 TeV data are based on an integrated luminosityof 4.7 fb − , while the 8 TeV data are based on an integrated luminosity of 20.3 fb − . We nowdiscuss each of these two datasets in turn.The 7 TeV data are taken at the Z peak, reaching values of the Z transverse momentumof up to p ZT = 800 GeV. They are given inclusively for Z/γ ∗ rapidities up to | y Z | = 2 .
4, aswell as in three separated rapidity bins given by 0 . ≤ | y Z | ≤ .
0, 1 . ≤ | y Z | ≤ . . ≤ | y Z | ≤ .
4. In order to maximize the potential constraint on PDFs, only the differentialmeasurement will be considered. The measurement is presented in terms of normalized cross-sections (1 /σ Z ) dσ ( Z ) /dp ZT , where σ Z is the fiducial cross-section in the corresponding di-leptonrapidity bin. This dataset has been left out of default NNPDF3.1 dataset, for reasons to bediscussed in Sect. 4.2.The 8 TeV dataset, which reaches p ZT values as high as 900 GeV, is presented in threeseparate invariant mass bins: low mass below the Z -peak, on-peak, and high mass above the15 -peak up to M ll = 150 GeV. In addition, the measurement taken at the Z -peak is providedboth inclusively in the whole rapidity range 0 . < | y Z | < . < y Z < .
4, 0 . < | y Z | < .
8, 0 . < | y Z | < .
2, 1 . < | y Z | < . . < | y Z | < . . < | y Z | < .
4. Once again, here the more differential measurementwill be used. In contrast to the 7 TeV data, the dataset is given both in terms of normalizedand absolute distributions. We will use the latter, not only because of the extra informationon the cross-section normalization, but also as problems can occur whenever the data used tocompute the normalization are provided in a range which differs from that of the data used forPDF determination. This problem is discussed in detail in Ref. [121] and described in Sect. 4.2.CMS has measured the cross-sections differentially in p T and rapidity y Z at 8 TeV [83], basedon an integrated luminosity of 19.7 fb − in the muon channel. Data is provided in five rapiditybins 0 . < | y Z | < .
4, 0 . < | y Z | < .
8, 0 . < | y Z | < .
2, 1 . < | y Z | < . . < | y Z | < . p ZT ≥
30 GeV (resummation would be requiredfor smaller p T ) [121]. Secondly, removing regions in which electroweak corrections are large andcomparable to the experimental data imposes a cut of p ZT ≤
150 (170) GeV for the ATLAS (CMS)data [121]. Finally, the CMS dataset in the largest rapidity bin is discarded due to an apparentincompatibility with both the corresponding ATLAS measurement in the same bin and thetheoretical prediction. The origin of this incompatibility remains unclear [121].Theoretical predictions have been obtained from Ref. [121], based upon the NNLO computa-tion of Z +jet production of Refs. [119,120]. Factorization and renormalization scales are chosenas µ R = µ F = (cid:113) ( p T ) + M ll , (2.2)where M ll is the invariant mass of the final-state lepton pair. The calculation includes the Z and γ ∗ contributions, their interference and decay to lepton pairs. The NNLO/NLO ratio isshown in Fig. 2.5 for the observables with the ATLAS and CMS acceptance cuts, computedusing NNPDF3.0 PDFs, with α s ( m Z ) = 0 . p T up to around 10% at high p T and is therefore required in order to describe data withsub-percent accuracy.Even the most accurate results for the NNLO/NLO correction factor still display fluctuations,as shown in Fig. 2.5 where we plot the NNLO/NLO cross-section ratio for the central rapiditybin of the 8 TeV ATLAS data. The points are shown together with their nominal MonteCarlo integration uncertainty [121]. The point-to-point statistical fluctuation of the theoreticalprediction appears to be larger than the typical uncorrelated statistical uncertainty on theATLAS dataset, which is typically at the sub-percent or even permille level. In order to checkthis, we have fitted an ensemble of neural networks to the cross-section ratio, as a function of p ZT for fixed rapidity. The fit has been performed in each of the rapidity bins for the ATLASand CMS data; more details are given in Ref. [123]. The result of the fit and its one-sigmauncertainty are shown in Fig. 2.6 for the central rapidity bin of the ATLAS data.The one-sigma uncertainty of the fit, which is determined by the point-to-point fluctuationof the NNLO computation, is at the percent level, which is rather larger than the statisticaluncertainty of the data. Indeed, it is clear by inspection of Figs. 2.5-2.6 that the point-to-pointfluctuations of the NNLO/NLO ratio are much larger than those of the data themselves (as seenin Ref.s [78], [121]). We conclude that there is a residual theoretical uncertainty on the NNLOprediction which we estimate to be of order of 1% for all datasets. This conclusion has beenvalidated and cross-checked by repeating the fit with cuts or different functional forms. We have16 σ d σ /dy Z /dM ll )ATLAS, 7 TeV LHC66 GeV < M ll < 116 GeVp T Z < 1.01.0 < Y Z < 2.02.0 < Y Z < 2.5 Z < 0.40.4 < Y Z < 0.80.8 < Y Z < 1.2 0.9 1 1.1 1.2 40 50 100 200 400K-factors (d σ /dy Z /dM ll )CMS, 8 TeV LHC66 GeV < M ll < 116 GeVp T Z < 1.61.6 < Y Z < 2.0 Z < 0.40.4 < Y Z < 0.80.8 < Y Z < 1.2 0.9 1 1.1 1.2 40 50 100 200 400K-factors (d σ /dy Z /dM ll )ATLAS, 8 TeV LHC66 GeV < M ll < 116 GeVp T Z < 1.61.6 < Y Z < 2.02.0 < Y Z < 2.4 ll < 20 GeV20 GeV < M ll < 30 GeV30 GeV < M ll < 46 GeV 0.9 1 1.1 1.2 40 50 100 200 400K-factors (d σ /dy Z /dM ll )ATLAS, 8 TeV LHC0 < Y Z < 2.4p T
46 GeV < M ll < 66 GeV116 GeV < M ll < 160 GeV Figure 2.5:
The NNLO/NLO cross-section for the
Z p T data corresponding to the acceptance cuts andbinning of the ATLAS 7 TeV (top left), CMS 8 TeV (top right), and the ATLAS 8 TeV (bottom) rapidity(left) and invariant mass (right) distributions. therefore added an extra 1% fully uncorrelated theoretical uncertainty to this dataset (see alsoRef. [121]). t ¯ t production Differential distributions for top pair production have been included in NNPDF3.1 followingthe detailed study of Ref. [124]. ATLAS and CMS have performed measurements of these dis-tributions with a variety of choices of kinematic variables, including the top quark rapidity y t ,the rapidity of the top pair y t ¯ t , the transverse momentum of the top quark p tT , and the invari-ant mass of the top-antitop system m t ¯ t . For ATLAS both absolute and normalized differentialdistributions are provided, whereas CMS only provides normalized results. Perturbative QCDcorrections for all these distributions have been computed at NNLO [15, 16]. In order to avoiddouble counting, only one distribution per experiment can be included in the dataset, as thestatistical correlations between different distributions are not available. The choice of differ-ential distributions adopted in NNPDF3.1 follows the recommendation of Ref. [124], where acomprehensive study of the impact on the gluon PDF of various combinations of differentialtop pair distributions was performed. It was found that the normalized rapidity distributionshave the largest constraining power and lead to a good agreement between theory and data17igure 2.6: The NNLO/NLO cross-section ratio in the central rapidity bin of the 8 TeV ATLAS
Z p T distribution. The result of a fit and its associate uncertainty are also shown. for ATLAS and CMS. The use of rapidity distributions has some further advantages. First, itreduces the risk of possible contamination by BSM effects. For example, heavy resonances wouldbe kinematically suppressed in the rapidity distributions, but not in the tails of the m t ¯ t and p tT distributions. Second, rapidity distributions exhibit a milder sensitivity upon variations of thevalue of m t than the p tT and m t ¯ t distributions [125].We therefore include the 8 TeV normalized rapidity distributions in the lepton+jets final statefrom ATLAS [73] and CMS [81], which correspond respectively to an integrated luminosity of20 . − and 19 . − . We consider measurements in the full phase space, with observablesreconstructed in terms of the top or top-pair kinematic variables, because NNLO results areavailable only for stable top quarks. We also include, again following Ref. [124], the mostrecent total cross-sections measurements at 7, 8 and 13 TeV from ATLAS [74, 75] and CMS [82,88]. They replace previous measurements from ATLAS [61–63] and CMS [64–66] included inNNPDF3.0.At NLO theoretical predictions have been generated with Sherpa [111], in a format compliantto
APPLgrid [106], using the
MCgrid code [112] and the
Rivet [126] analysis package, with
OpenLoops [127] for the NLO matrix elements. All calculations have been performed with largeMonte Carlo integration statistics in order to ensure that residual numerical fluctuations arenegligible. Our results have been carefully benchmarked against those obtained from the codeof [16]. Renormalization and factorization scales, µ R and µ F respectively, have been chosenbased on the recommendation of Ref. [16] as µ R = µ F = µ = H T / , H T ≡ (cid:113) m t + (cid:0) p tT (cid:1) + (cid:113) m t + (cid:0) p ¯ tT (cid:1) , (2.3)where m t = 173 . p tT ( p ¯ tT ) is the top (anti-top) transverse momentum. NLO theoretical predictions for normalizeddifferential distributions have been obtained by dividing their absolute counterparts by thecross-section integrated over the kinematic range of the data.The NNLO correction factors have been computed separately for the absolute differentialcross-sections and their normalizing total cross-sections. Differential cross-sections have beendetermined using the code of [16], with the scale choice Eq. (2.3). Results for the NNLO/NLOratio are shown in Fig. 2.7, where it can be seen that the size of the NNLO corrections is 6% and9%, actually smaller than the data uncertainty, with a reasonably flat shape in the kinematicregion covered by the data. We also show explicitly the dependence of the results on the PDF setused in the calculation by using three different global PDF sets: it is clear that this dependenceis completely negligible.Total cross-sections have been computed with the top++ code [129] at NNLO+NNLL, andwith fixed scales µ R = µ F = m t , following the recommendation of Ref. [16] which suggests that18 - . - . - . - . - . . . . . . y t K factor (d σ /dy t ) NNPDF3.0CT14MMHT14 - . - . - . - . - . . . . . . y tt- K factor (d σ /dy tt- ) NNPDF3.0CT14MMHT14 Figure 2.7:
The NNLO/NLO cross-section ratio for the top quark rapidity y t (left) and top-quark pairrapidity y t ¯ t (right) corresponding to the 8 TeV ATLAS and CMS data. Results obtained with threedifferent input PDF sets, NNPDF3.0, CT14, and MMHT14, are shown. NNLO+NNLL resummed cross-sections should be used in conjunction to NNLO differentialdistributions if the latter are determined using a dynamical scale choice.19
The NNPDF3.1 global analysis
We now present the results of the NNPDF3.1 global analysis at LO, NLO and NNLO, andcompare them with the previous release NNPDF3.0 and with other recent PDF sets. Here wepresent results obtained using the complete dataset of Tab. 2.1-2.3, discussed in Sect. 2. Studiesof the impact of individual measurements will be discussed along with PDF determinations fromreduced datasets in Sect. 4.After a brief methodological summary, we discuss the fit quality, and then examine individ-ual PDFs and their uncertainties. We compare NNPDF3.1 PDFs with NNPDF3.0 and withCT14 [6], MMHT2014 [7] and ABMP16 [8]. We next examine the impact of independentlyparametrizing charm, the principal methodological improvement in NNPDF3.1. Finally, wediscuss theoretical uncertainties, both related to QCD parameters and to missing higher ordercorrections to the theory used for PDF determination.In this Section all NLO and NNLO NNPDF3.1 results are produced using the CMC [25]optimized 100 replica Monte Carlo sets, see Sect. 6.2 below: despite only including 100 replicas,these sets reproduce the statistical features of a set of at least about 400 replicas (see Sect. 6.1).We present here only a selection of results: a more extensive set of results is available from apublic repository, see Sect. 6.2.
NNPDF3.1 PDFs are determined with largely the same methodology as in NNPDF3.0: the onlysignificant change is that now charm is independently parametrized. The PDF parametrizationis identical to that discussed in Sect. 3.2 of Ref. [5], including the treatment of preprocessing, butwith the PDF basis in Eq. (3.4) of that reference now supplemented by an extra PDF for charm,parametrized like all other PDFs (as per Eq. (2) of Ref. [23]). PDFs are parametrized at the scale Q = 1 .
65 GeV whenever the charm PDF is independently parametrized. For the purposes ofcomparison we also provide PDF sets constructed with perturbatively generated charm; in thesesets, PDFs are parametrized at the scale Q = 1 . α s ( m Z ) = 0 .
118 as a default throughout the paper, though determi-nations have also been performed for several different values of α s (see Sect. 6.2). Pole heavyquark masses are used throughout, with the main motivation that for the inclusive observablesused for PDF determination MS masses are inappropriate, since they distort the perturbativeexpansion in the threshold region [130]. The default values of the heavy quark pole masses are m c = 1 .
51 GeV for charm and m b = 4 .
92 GeV for bottom, following the recommendation of theHiggs cross-section working group [131]; PDF sets for different charm mass values, correspondingto the ± In Table 3.1 we provide values of χ /N dat both for the global fit and individually for all thedatasets included in the NNPDF3.1 LO, NLO and NNLO PDF determinations. These arecompared to their NNPDF3.0 NLO and NNLO counterparts. The χ is computed using thecovariance matrix including all correlations, as published by the corresponding experiments.Inspection of this table shows that the fit quality improves from LO to NLO to NNLO: not onlyis there a significant improvement between LO and NLO, but there is also a marked improvementwhen going from NLO to NNLO. It is interesting to note that this was not the case in NNPDF3.0where the fit quality at NNLO was in fact slightly worse than at NLO (see Table 9 of Ref. [5]).20his reflects the increased proportion of hadronic processes included in NNPDF3.1, for whichNNLO corrections are often substantial, and also, possibly, methodological improvements.The overall fit quality with NNPDF3.1 is rather better than that obtained using NNPDF3.0PDFs. Whereas this is clearly expected for LHC measurements which were not included inNNPDF3.0, it is interesting to note that the HERA measurements which were already presentin 3.0 (though in slightly different uncombined form) are also better fitted. The quality ofthe description with the previous NNPDF3.0 PDFs is nevertheless quite acceptable for all thenew data, indicating a general compatibility between NNPDF3.0 and NNPDF3.1. Note thatNNPDF3.0 values in Table 3.1 are computed using the NNPDF3.1 theory settings, thus inparticular with different values of the heavy quark masses than those used in the NNPDF3.0PDF determination. Because of this, the NNPDF3.0 fit quality shown shown in Table 9 ofRef. [5] is slightly better than that shown in Table 3.1, yet even so the fit quality of NNPDF3.1is better still. Specifically, concerning HERA data, the fit quality of NNPDF3.0 with consistenttheory settings can be read off Table 7 of Ref. [124]: it corresponds to χ /N dat = 1 .
21 therebyshowing that indeed NNPDF3.1 provides a better description. The reasons for this improvementwill be discussed in Sect. 3.4 below.For many of the new LHC measurements, achieving a good description of the data is onlypossible at NNLO. The total χ /N dat for the ATLAS, CMS and LHCb experiments is 1.09,1.06 and 1.47 respectively at NNLO, compared to 1.36, 1.20 and 1.62 at NLO. The datasetsexhibiting the largest improvement when going from NLO to NNLO are those with the smallestexperimental uncertainties. For example the ATLAS W, Z
Z p T distributions (from 3.65 to 1.32) and the LHCb 8 TeV W, Z → µ rapidity distributions (from 1.88 to 1.37); in these experiments uncorrelated statisticaluncertainties are typically at the sub-percent level. It is likely that this trend will continue asLHC measurements become more precise. 21 NPDF3.1 NNPDF3.0Dataset NNLO NLO LO NNLO NLONMC 1.30 1.35 3.25 1.29 1.36SLAC 0.75 1.17 3.35 0.66 1.08BCDMS 1.21 1.17 2.20 1.31 1.21CHORUS 1.11 1.06 1.16 1.11 1.14NuTeV dimuon 0.82 0.87 4.75 0.69 0.61HERA I+II inclusive 1.16 1.14 1.77 1.25 1.20HERA σ NC c F b σ d DY /σ p DY σ p σ p Z rap 1.48 1.619 1.54 1.55 1.28CDF Run II k t jets 0.87 0.84 1.07 0.82 0.95D0 Z rap 0.60 0.67 0.65 0.61 0.59D0 W → eν asy 2.70 1.59 1.75 [2.68] [4.58]D0 W → µν asy 1.56 1.52 2.16 [2.02] [1.43]ATLAS total ATLAS
W, Z
W, Z
Z p T p llT , M ll ) 0.93 1.17 - [1.05] [1.28]ATLAS Z p T p llT , y ll ) 0.94 1.77 - [1.19] [2.49]ATLAS σ tot tt t ¯ t rap 1.45 1.31 1.99 [3.32] [1.50]CMS total CMS W asy 840 pb 0.78 0.86 1.55 0.73 0.85CMS W asy 4.7 fb 1.75 1.77 3.16 1.75 1.82CMS W + c tot - 0.54 16.5 - 0.93CMS W + c ratio - 1.91 3.21 - 2.09CMS Drell-Yan 2D 2011 1.27 1.23 2.15 1.20 1.19CMS W rap 8 TeV 1.01 0.70 4.32 [1.24] [0.96]CMS jets 7 TeV 2011 0.84 0.84 0.93 1.06 0.98CMS jets 2.76 TeV 1.03 1.01 1.09 [1.22] [1.18]CMS Z p T p llT , M ll ) 1.32 3.65 - [1.59] [3.86]CMS σ tot tt t ¯ t rap 0.94 0.96 1.32 [1.15] [1.01]LHCb total LHCb Z
940 pb 1.49 1.27 2.51 1.29 0.91LHCb Z → ee W, Z → µ W, Z → µ Total dataset 1.148 1.168 2.238 1.284 1.307
Table 3.1:
The values of χ /N dat for the global fit and for all the datasets included in the NNPDF3.1LO, NLO and NNLO PDF determinations. Values obtained using the NNPDF3.0 NLO and NNLO PDFsare also shown: numbers in brackets correspond to data not fitted in NNPDF3.0. Note that NNPDF3.0values are produced using NNPDF3.1 theory settings, and are thus somewhat worse than those quotedin Ref. [5]. - - -
10 100.10.20.30.40.50.60.70.80.91 g/10 v u v ddcs u NNPDF3.1 (NNLO) ) =10 GeV m xf(x, x - - -
10 100.10.20.30.40.50.60.70.80.91 g/10 v u v dd us cb ) GeV =10 m xf(x, x - - -
10 100.10.20.30.40.50.60.70.80.91 g/10 v u v ddcs u NNPDF3.1 (NNLO) ) =10 GeV m xf(x, x - - -
10 100.10.20.30.40.50.60.70.80.91 g/10 v u v dd us cb ) GeV =10 m xf(x, Figure 3.1:
The NNPDF3.1 NNLO PDFs, evaluated at µ = 10 GeV (left) and µ = 10 GeV (right). We now inspect the baseline NNPDF3.1 parton distributions, and compare them to NNPDF3.0and to MMHT14 [7], CT14 [6] and ABMP16 [8]. The NNLO NNPDF3.1 PDFs are displayedin Fig. 3.1. It can be seen that although charm is now independently parametrized, it is stillknown more precisely than the strange PDF. The most precisely determined PDF over most ofthe experimentally accessible range of x is now the gluon, as will be discussed in more detailbelow.In Fig. 3.2 we show the distance between the NNPDF3.1 and NNPDF3.0 PDFs. Accordingto the definition of the distance given in Ref. [98], d (cid:39) N rep = 100 replicas, a distance of d (cid:39)
10 corresponds to adifference of one-sigma in units of the corresponding variance, both for central values and forPDF uncertainties. For clarity only the distance between the total strangeness distributions s + = s + ¯ s is shown, rather than the strange and antistrange separately. We find importantdifferences both at the level of central values and of PDF errors for all flavors and in the entirerange of x . The largest distance is found for charm, which is independently parametrized inNNPDF3.1, while it was not in NNPDF3.0. Aside from this, the most significant distances areseen in light quark distributions at large x and strangeness at medium x .In Fig. 3.3 we compare the full set of NNPDF3.1 NNLO PDFs with NNPDF3.0. TheNNPDF3.1 gluon is slightly larger than its NNPDF3.0 counterpart in the x ∼ < .
03 region, whileit becomes smaller at larger x , with significantly reduced PDF errors. The NNPDF3.1 lightquarks and strangeness are larger than 3.0 at intermediate x , with the largest deviation seenfor the strange and antidown PDFs, while at both small and large x there is good agreementbetween the two PDF determinations. The best-fit charm PDF of NNPDF3.1 is significantly23 − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO vs NNPDF3.0 NNLO
Figure 3.2:
Distances between the central values (left) and the uncertainties (right) of the NNPDF3.0and NNPDF3.1 NNLO PDF sets, evaluated at Q = 100 GeV. Note the different in scale on the y axisbetween the two plots. smaller in the intermediate- x region compared to the perturbative charm of NNPDF3.0, whileat larger x it has significantly increased uncertainty.A detailed comparison of the corresponding uncertainties is presented in Fig. 3.4, wherewe compare the relative uncertainty on each PDF, defined as the ratio of the one-sigma PDFuncertainty to the central value of the NNPDF3.1 set. NNPDF3.1 uncertainties are eithercomparable to those of NNPDF3.0, or are rather smaller. The only major exception to this isthe charm PDF at intermediate and large x for which uncertainties are substantially increased.On the other hand, the uncertainties in the gluon PDF are smaller in NNPDF3.1 over theentire range of x . This is an important result, since one may have expected generally largeruncertainties in NNPDF3.1 due to the inclusion of one additional freely parametrized PDF. Thefact that the only uncertainty which has enlarged significantly is that of the charm PDF suggeststhat not parametrizing charm may be a source of bias. The fact that central values changeby a non-negligible amount, though compatible within uncertainties, while the uncertaintiesthemselves are significantly reduced, strongly suggests that NNPDF3.1 is more accurate thanNNPDF3.0, as would be expected from the substantial amount of new data included in the fit.The effect of parametrizing charm on PDFs and their uncertainties will be discussed in moredetail in Sect. 3.4, while the effects of the new data on both central values and uncertainties willbe discussed in Sect. 4.1.In Fig. 3.5 we compare the NNPDF3.1 PDFs to the other global PDF sets included in thePDF4LHC15 combination along with NNPDF3.0, namely CT14 and MMHT14. This compari-son is therefore indicative of the effect of replacing NNPDF3.0 with NNPDF3.1 in the combina-tion. The relative uncertainties in the three sets are compared in Fig. 3.6. Comparing Fig. 3.5to Fig. 3.3, it is interesting to observe that several aspects of the pattern of differences betweenNNPDF3.1 and the other global fits are similar to those between NNPDF3.1 and NNPDF3.0,and therefore they are likely to have a similar origin. This is patricularly clear for the charm andgluon. The gluon in the region x ∼ < .
03, relevant for Higgs production, is still in good agree-ment between the three sets. However, now NNPDF3.1 is at the upper edge of the one-sigmarange, i.e. the NNPDF3.1 gluon in this region is enhanced. At large x the NNPDF3.1 gluonis instead suppressed in comparison to MMHT14 and CT14. As we will show in Sects. 3.4, 4.1the enhancement is a consequence of parametrizing charm, while as we will show in Sect. 4.324he large- x suppression is a direct consequence of including the 8 TeV top differential data. Theuncertainty in the NNPDF3.1 gluon PDF is now noticeably smaller than that of either CT14 orMMHT14.For the quark PDFs, for up and down we find good agreement in the entire range of x .For the antidown PDF, agreement is marginal, with NNPDF3.1 above MMHT14 and CT14 for x ∼ < . x . The strange fraction of the proton is larger in NNPDF3.1than CT14 and MMHT14, and has rather smaller PDF uncertainties. The best-fit NNPDF3.1charm is suppressed at intermediate x in comparison to the perturbatively generated ones ofCT14 and MMHT14, but has a much larger uncertainty at large x as would be expected, withthe differences clearly traceable to the fact that in NNPDF3.1 charm is freely parametrized.Finally, in Fig. 3.7 we compare NNPDF3.1 to the recent ABMP16 set. This set is releasedin various fixed-flavor number schemes. Because we perform the comparison at a scale Q = 10 GeV , we choose the n f = 5 NNLO ABMP16 sets, both with their default value α s ( m Z ) = 0 . α s ( m Z ) = 0 . α s ( m Z ) = 0 .
118 is adopted, there isgenerally reasonable agreement for the gluon PDF, except at large x where ABMP16 undershootsNNPDF3.1. Differences are larger in the case of light quarks: the ABMP16 up distributionovershoots NNPDF3.1 at large x , while the down quark undershoots in the whole x range.Differences are largest for the strange PDF, though comparing with Fig. 3.5 it is clear thatABMP16 differs by a similary large amount from MMHT14 and CT14. In general the ABPM16sets have rather smaller uncertainties than NNPDF3.1. This is especially striking for strangeness,where the difference in uncertainty is particularly evident. This is to be contrasted with the CT14and MMHT14 sets, which have qualitatively similar uncertainties to NNPDF3.1 throughoutthe data region. The fact that the uncertainties for ABMP16 are so small can be traced totheir overly restrictive parametrization, and the fact that this set is produced using a Hessianmethodology, but unlike MMHT14 and CT14, with no tolerance (see Refs. [6, 7]).25 x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q u ) / ( x , Q u NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q d ) / ( x , Q d NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / s ( x , Q s ( x , Q NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q s ) / ( x , Q s NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q + ) / c ( x , Q + c NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV
Figure 3.3:
Comparison between NNPDF3.1 and NNPDF3.0 NNLO PDFs at Q = 100 GeV. From topto bottom up and antiup, down and antidown, strange and antistrange, charm and gluon are shown. x - - - - ) [ r e f] ) ) / ( u ( x , Q u ( x , Q d NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q u ) / ( ( x , Q u d NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( d ( x , Q d ( x , Q d NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q d ) / ( ( x , Q d d NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( s ( x , Q s ( x , Q d NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q s ) / ( ( x , Q s d NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q + ) / ( c ( x , Q + c d NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( g ( x , Q g ( x , Q d NNPDF3.1NNPDF3.0
NNLO, Q = 100 GeV
Figure 3.4:
Comparison between NNPDF3.1 and NNPDF3.0 relative PDF uncertainties at Q = 100;the PDFs are as in Fig. 3.3. The uncertainties shown are all normalized to the NNPDF3.1 central value. x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q u ) / ( x , Q u NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q d ) / ( x , Q d NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / s ( x , Q s ( x , Q NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q s ) / ( x , Q s NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q + ) / c ( x , Q + c NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV
Figure 3.5:
Comparison between NNPDF3.1, CT14 and MMHT2014 NNLO PDFs. The comparison isperformed at Q = 100 GeV, and results are shown normalized to the central value of NNPDF3.1; thePDFs are as in Fig. 3.3. x - - - - ) [ r e f] ) ) / ( u ( x , Q u ( x , Q d NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q u ) / ( ( x , Q u d NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( d ( x , Q d ( x , Q d NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q d ) / ( ( x , Q d d NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( s ( x , Q s ( x , Q d NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q s ) / ( ( x , Q s d NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q + ) / ( c ( x , Q + c d NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( g ( x , Q g ( x , Q d NNPDF3.1CT14MMHT2014
NNLO, Q = 100 GeV
Figure 3.6:
Comparison between NNPDF3.1, CT14 and MMHT2014 relative PDF uncertainties at Q = 100; the PDFs are as in Fig. 3.5. x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q NNPDF3.1 =0.118 S a ABMP16 =0.1147 S a ABMP16
NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q u ) / ( x , Q u NNPDF3.1 =0.118 S a ABMP16 =0.1147 S a ABMP16
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q NNPDF3.1 =0.118 S a ABMP16 =0.1147 S a ABMP16
NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q d ) / ( x , Q d NNPDF3.1 =0.118 S a ABMP16 =0.1147 S a ABMP16
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / s ( x , Q s ( x , Q NNPDF3.1 =0.118 S a ABMP16 =0.1147 S a ABMP16
NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q s ) / ( x , Q s NNPDF3.1 =0.118 S a ABMP16 =0.1147 S a ABMP16
NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q + ) / c ( x , Q + c NNPDF3.1 =0.118 S a ABMP16 =0.1147 S a ABMP16
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNPDF3.1 =0.118 S a ABMP16 =0.1147 S a ABMP16
NNLO, Q = 100 GeV
Figure 3.7:
Same as Fig. 3.5 but now comparing to the ABMP16 NNLO n f = 5 sets both with theirdefault α s ( m Z ) = 0 . α s ( m Z ) = 0 . .4 Methodological improvements: parametrizing charm The main methodological improvement in NNPDF3.1 over NNPDF3.0 is the fact that the charmPDF is now parametrized in the same way as the light and strange quark PDFs. To quantifythe effect of this change, we have performed a repeat of the NNPDF3.1 analysis but with charmtreated as in all previous NNPDF PDF determinations, i.e., generated entirely perturbativelythrough matching conditions implemented at NLO or NNLO.In Table 3.2 we show the χ /N dat values when charm is perturbatively generated at NLO andNNLO. Unsurprisingly the fit quality deteriorates when charm is not independently parametrizedPDF. This is what one would naively expect since perturbative charm imposes a constraint uponthe fit, thereby reducing the number of free parameters.However, it is interesting to observe that the fit quality to the inclusive HERA data (1306data points) significantly deteriorates when going from NLO to NNLO with perturbative charm,whereas it remains stable when charm is independently parametrized. Concerning the charmstructure function data, note that, as discussed in Sect. 2.2 above, a further cut is applied to theHERA σ NC c data at NNLO when charm is independently parametrized. In order to allow for aconsistent comparison, in Table 3.2 we show in parenthesis the value of χ /N dat computed forthe 37 (out of 47) data points that survive this cut also for all other cases. Hence, for this datathe fit quality is simlar with perturbative and parametrized charm, and also similar at NLO andNNLO (slightly worse at NNLO, by an amount compatible with a statistical fluctuation). Thefact that when parametrizing charm there no longer is a deterioration of fit quality when goingfrom NLO to NNLO suggests that this resolves a tension present at NNLO, with perturbativecharm, between HERA and hadron collider data. Likewise, a purely perturbative charm leads toa substantial deterioration at NNLO for BCDMS, NMC and especially for the NuTeV dimuoncross-sections. This can be traced to the fact that independently parametrizing charm is essentialto reconcile the HERA data with the constraints on the strange content of the proton imposedby the ATLAS W, Z x ∼ > .
003 and reduced for smaller x when charm is independently parametrized. The largest differences can be seen in the up quark,while the strange and gluon distributions are more stable. The best-fit charm distribution has adistinctly different shape and significantly larger uncertainty than its perturbatively generatedcounterpart. As argued in Ref. [23] this shape might well be compatible with a charm PDFgenerated perturbatively at high perturbative orders.In Fig. 3.9 we directly compare PDF uncertainties. It is remarkable that the uncertaintiesother than for charm are essentially unchanged when charm is independently parametrized, withonly a slight increase in sea quark PDF uncertainties for 10 − ∼ < x ∼ < − . The uncertainty onthe gluon is almost completely unaffected. The PDF uncertainty on charm when it is indepen-dently parametrized is in line with that of other sea quark PDFs, while the uncertainty of theperturbatively generated charm follows that of the gluon and is consequently much smaller.A previous comparison of PDFs determined with parametrized or perturbative charm waspresented in Ref. [23] and led to the conclusion that parametrizing charm and determining itfrom the data greatly reduces the dependence on the charm mass thereby reducing the overallPDF uncertainty when the uncertainty due to the charm mass is kept into account. As men-tioned, NNPDF3.1 PDFs are determined using heavy quark pole mass values and uncertaintiesrecommended by the Higgs Cross-Section Working Group [131]. For charm, this corresponds to m pole c = 1 . ± .
13 GeV. In order to estimate the impact of this uncertainty, we have producedNNPDF3.1 NNLO sets with m pole c = 1 .
38 GeV and m pole c = 1 .
64 GeV. Results are shown inFig. 3.10 for some representative PDFs, both for the default NNPDF3.1 and for the versionwith perturbative charm. It is clear that the very strong dependence of the charm PDF on31 c which is found when charm is perturbatively generated all but disappears when charm isindependently parametrized. While the gluon is always quite stable, the dependence of pertur-batively generated charm on m c propagates to the light quark distributions. These are thereforesignificantly stabilized by parametrizing charm. Indeed, if charm is generated perturbatively,the shift in up and down quark distribution upon one-sigma variation of the charm mass iscomparable to (though somewhat smaller than) the PDF uncertainty. When charm is inde-pendently parametrized this dependence is considerably reduced. With parametrized charm,collider observables at high scales become essentially independent of the charm mass, in linewith the expectation from decoupling arguments.32 NPDF3.1 pert. charm NNPDF3.1Dataset NNLO NLO NNLO NLONMC 1.38 1.38 1.30 1.35SLAC 0.70 1.22 0.75 1.17BCDMS 1.27 1.24 1.21 1.17CHORUS 1.10 1.07 1.11 1.06NuTeV dimuon 1.27 1.01 0.82 0.87HERA I+II inclusive 1.21 1.15 1.16 1.14HERA σ NC c F b σ d DY /σ p DY σ p σ p Z rap 1.44 1.46 1.48 1.62CDF Run II k t jets 0.86 0.86 0.87 0.84D0 Z rap 0.60 0.64 0.60 0.67D0 W → eν asy 2.71 1.63 2.70 1.59D0 W → µν asy 1.42 1.38 1.56 1.52ATLAS total ATLAS
W, Z
W, Z
Z p T p llT , M ll ) 0.94 1.19 0.93 1.17ATLAS Z p T p llT , y ll ) 0.96 1.84 0.94 1.77ATLAS σ tot tt t ¯ t rap 1.39 1.18 1.45 1.31CMS total CMS W asy 840 pb 0.69 0.80 0.78 0.86CMS W asy 4.7 fb 1.75 1.76 1.75 1.77CMS W + c tot - 0.49 - 0.54CMS W + c ratio - 1.92 - 1.91CMS Drell-Yan 2D 2011 1.33 1.27 1.27 1.23CMS W rap 8 TeV 0.90 0.65 1.01 0.70CMS jets 7 TeV 2011 0.87 0.86 0.84 0.84CMS jets 2.76 TeV 1.06 1.05 1.03 1.01CMS Z p T p llT , y ll ) 1.29 3.50 1.32 3.65CMS σ tot tt t ¯ t rap 0.96 0.96 0.94 0.96LHCb total LHCb Z
940 pb 1.31 1.08 1.49 1.27LHCb Z → ee W, Z → µ W, Z → µ Table 3.2:
Same as Tab. 3.1, but now comparing the default NNPDF3.1 NNLO and NNLO sets to thevariant in which charm is perturbatively generated. For HERA σ NC c the number in parenthesis refer tothe subset of data to which the NNLO FC cut of Table. 2.4 is applied. x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q Fitted charmPerturbative charmNNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q Fitted charmPerturbative charmNNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q u ) / ( x , Q u Fitted charmPerturbative charmNNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q d ) / ( x , Q d Fitted charmPerturbative charmNNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / s ( x , Q s ( x , Q Fitted charmPerturbative charm
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q s ) / ( x , Q s Fitted charmPerturbative charm
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q + ) / c ( x , Q + c Fitted charmPerturbative charmNNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q Fitted charmPerturbative charmNNPDF3.1 NNLO, Q = 100 GeV
Figure 3.8:
Comparison of NNPDF3.1 NNLO PDFs to a variant in which charm is generated entirelyperturbatively (and everything else is unchanged). x - - - - ) [ r e f] ) ) / ( u ( x , Q u ( x , Q d NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( d ( x , Q d ( x , Q d NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q u ) / ( ( x , Q u d NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q d ) / ( ( x , Q d d NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( s ( x , Q s ( x , Q d NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q s ) / ( ( x , Q s d NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q + ) / ( c ( x , Q + c d NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( g ( x , Q g ( x , Q d NNPDF3.1 NNLO, Q = 100 GeV
Figure 3.9:
Comparison of the fractional one-sigma PDF uncertainties in NNPDF3.1 NNLO with thecorresponding version where charm is generated perturbatively (and everything else is unchanged). ThePDF comparison plot was shown in Fig. 3.8. x - - - - ) [ r e f] ( x , Q + ) / c ( x , Q + c =1.51 GeV c m =1.38 GeV c m =1.64 GeV c m NNPDF3.1 NNLO fitted charm, Q = 100 GeV x - - - - ) [ r e f] ( x , Q + ) / c ( x , Q + c =1.51 GeV c m =1.38 GeV c m =1.64 GeV c m NNPDF3.1 NNLO perturbative charm, Q = 100 GeV x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q =1.51 GeV c m =1.38 GeV c m =1.64 GeV c m NNPDF3.1 NNLO fitted charm, Q = 100 GeV x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q =1.51 GeV c m =1.38 GeV c m =1.64 GeV c m NNPDF3.1 NNLO perturbative charm, Q = 100 GeV x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q =1.51 GeV c m =1.38 GeV c m =1.64 GeV c m NNPDF3.1 NNLO fitted charm, Q = 100 GeV x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q =1.51 GeV c m =1.38 GeV c m =1.64 GeV c m NNPDF3.1 NNLO perturbative charm, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q =1.51 GeV c m =1.38 GeV c m =1.64 GeV c m NNPDF3.1 NNLO fitted charm, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q =1.51 GeV c m =1.38 GeV c m =1.64 GeV c m NNPDF3.1 NNLO perturbative charm, Q = 100 GeV
Figure 3.10:
Dependence of the NNPDF3.1 NNLO PDFs on the charm mass. Results are shown bothfor parametrized charm (left) and perturbative charm (right), for (from top to bottom) charm, gluon, upand down PDFs. .5 Theoretical uncertainties PDF uncertainties on global PDF sets entering the PDF4LHC15 combination consist only of theuncertainty propagated from experimental data and uncertainties due to the methodology. Thesecan be controlled through closure testing. There are however further sources of uncertainty dueto the theory used in PDF determination, which we briefly assess here. These can be dividedinto two main classes: • Missing higher order uncertainties (MHOU), arising due to the truncation of the QCDperturbative expansion at a given fixed order (LO, NLO or NNLO) in the theory used forPDF determination. • Parametric uncertainties, due to the uncertainties on the values of parameters of the theoryused for PDF determination: the main ones are the values of α s ( m Z ) and of m pole c .A full assessment of MHOU is an open problem, which we leave to future investigations.For the time being, a first assessment can be obtained by studying the perturbative stabilityof our results. In Fig. 3.11 we show the distances at Q = 100 GeV between all the PDFs inthe LO and NLO sets, and in the NLO and NNLO sets. Some of the LO, NLO and NNLOPDFs are then compared directly in Fig. 3.12. Differences between the LO and NLO sets arevery large, both for central values and uncertainties, the latter being substantial at LO due tothe poor fit quality. The shift in quark PDFs can be as large as two sigma ( d (cid:39) x is completely different between LO and NLO due to the fact that the singularsmall- x behaviour of the quark to gluon splittings only starts at NLO, and due to the vanishingof gluon initiated DIS and DY processes at LO. On the other hand, when going from NLO toNNLO, PDF uncertainties are essentially unaffected. Central values are also reasonably stable:the largest shifts, in the large- x gluon and down quark and small- x gluon, remain at or belowthe one-sigma level.A quantitative estimate of the MHOU can be obtained by computing the shift between thecentral values of the NLO and NNLO NNPDF3.1 PDFs. The result is shown in Fig. 3.13 forsome PDF combinations. In the plot, the shift has been symmetrized, and is compared to theNLO standard PDF uncertainty. In the quark singlet Σ for x (cid:46) − the shift is larger thanthe PDF uncertainty, while it is smaller for individual flavors (as illustrated by the two quarkdistributions shown). This suggests that for individual quark flavors and the gluon at NNLO,MHOU can be reasonably neglected at the current level of precision. However, for particularcombinations (such as the singlet at small x ) it is unclear whether MHOU can be neglected evenat NNLO, given that at NLO they are larger than the PDF uncertainty.We finally turn to parametric uncertainties. As we have discussed in Sect. 3.4, the depen-dence of PDFs upon the charm mass is almost entirely removed by parametrizing charm. Thedependence on the b -quark mass is minor, except for the bottom PDFs themselves [94, 132].Therefore, the only significant residual parametric uncertainty is on the value of the strongcoupling. This uncertainty is routinely included along with the PDF uncertainty; in order todo this consistently, one needs PDF sets produced with different central values of α s (see e.g.Ref. [12]). We have determined NNPDF3.1 NLO and NNLO PDFs with α s ( m Z ) varied in therange 0 . ≤ α s ( m Z ) ≤ .
124 (see Sect. 6.2).In Fig. 3.14 we compare the up and gluon PDFs as α s ( m Z ) is varied by ∆ α s = ± .
002 aboutits central value. As is well known, the gluon is anti-correlated to α s ( m Z ) at small and medium x , but positively correlated to it at large x . The dependence on α s is rather milder for quarkPDFs, with positive correlation at small x , and very little dependence altogether at large x .37 − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 LO vs NLO − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NLO vs NNLO
Figure 3.11:
Distances between the LO and NLO (top) and the NLO and NNLO (bottom) NNPDF3.1NNLO PDFs at Q = 100 GeV. Note the difference in scale on the y axis between the two plots. x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNLONLOLO
NNPDF3.1, Q = 100 GeV x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q NNLONLOLO
NNPDF3.1, Q = 100 GeV x - - - - ) [ r e f] ( x , Q d ) / ( x , Q d NNLONLOLO
NNPDF3.1, Q = 100 GeV x - - - - ) [ r e f] ( x , Q + ) / s ( x , Q + s NNLONLOLO
NNPDF3.1, Q = 100 GeV
Figure 3.12:
Comparison between some of the LO, NLO and NNPDF3.1 NNLO PDFs: gluon and up(top), antidown and total strangeness (bottom). All results are shown at Q = 100 GeV, normalized tothe NNLO central value. x - - - - ) ( x , Q S D NLO PDF uncertaintiesTH error (NLO => NNLO shift)
NNPDF3.1, Q = 100 GeV x - - - - ) g ( x , Q D NLO PDF uncertaintiesTH error (NLO => NNLO shift)
NNPDF3.1, Q = 100 GeV x - - - - ) u ( x , Q D NLO PDF uncertaintiesTH error (NLO => NNLO shift)
NNPDF3.1, Q = 100 GeV x - - - - ) ( x , Q d D NLO PDF uncertaintiesTH error (NLO => NNLO shift)
NNPDF3.1, Q = 100 GeV
Figure 3.13:
Comparison between the NLO PDF uncertainties and the shift between the NLO and NNLOPDFs. All results are shown as ratios to the NLO PDFs, for Q = 100 GeV. The shift is symmetrized.We show results for the singlet, gluon (top); up and antidown (bottom) PDFs. x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q )=0.118 Z (m S a )=0.116 Z (m S a )=0.120 Z (m S a NNPDF3.1 NLO, Q = 100 GeV x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q )=0.118 Z (m S a )=0.116 Z (m S a )=0.120 Z (m S a NNPDF3.1 NLO, Q = 100 GeV x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q )=0.118 Z (m S a )=0.116 Z (m S a )=0.120 Z (m S a NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q )=0.118 Z (m S a )=0.116 Z (m S a )=0.120 Z (m S a NNPDF3.1 NNLO, Q = 100 GeV
Figure 3.14:
Dependence of NNPDF3.1 NLO (top) and NNLO (bottom) PDFs on the value of α s . Thegluon (left) and up quark (right) are shown at Q = 100 GeV, normalized to the central value. − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of new data
Figure 4.1:
Same as Fig. 3.2, but now comparing the NNPDF3.1 NNLO global PDFs to PDFs determinedusing exactly the same methodology but with the NNPDF3.0 dataset.
We now study the dependence of the NNPDF3.1 PDF set upon the experimental information onwhich it is based. Firstly we disentangle the effects of new data from the effects of methodologicalchanges. Then we systematically quantify the impact on PDFs of each new piece of experimentalinformation added in NNPDF3.1. Finally we discuss PDF determinations based on particulardata subsets; PDFs determined only from collider data (i.e. excluding all fixed target data),only from proton data (i.e. excluding all nuclear data), or excluding all LHC data. As thesePDF sets based on reduced dataset can also be useful for specific phenomenological applications,they are also made available (see Section 6.2 below). As in the previous Section, here we willonly present a selection of representative plots, the interested reader is referred to a much largerset of plots available online as discussed in Section 6.2.
In Section 3.4 we have studied the impact of the main methodological improvement introduced inNNPDF3.1, namely, independently parametrizing the charm PDF and determining it from thedata. In order to completely disentangle the effect of data and methodology we have performed aPDF determination using NNPDF3.1 methodology, but the NNPDF3.0 dataset: specifically, wehave removed from the NNPDF3.1 dataset all the new data. There remain some small residualdifferences between this restricted dataset and that of NNPDF3.0, specifically in some smalldifferences in cuts and in the use of the combined HERA data instead of the separate HERA-Iand HERA-II sets. However, these differences are expected to be minor [93].In Fig. 4.1 we show the distances between the NNPDF3.1 NNLO PDF set, and that basedon the NNPDF3.0 dataset using the same methodology. We see that the impact of the newdata is mostly localized at large x , for the up, down and charm quarks and the gluon, and atmedium x for strangeness. As far as uncertainties are concerned, we observe improvements ofup to half a sigma across a wide range in x and for all PDF flavors. In Fig. 4.2 we comparesome representative PDFs for NNPDF3.1, the set based on NNPDF3.0 data with NNPDF3.1methodology, and the original NNPDF3.0. We see that the overall effect of the new data and thenew methodology are comparable, but that they act in different regions and for different PDFs.41or instance, for the light quarks and the gluon the impact of the new methodology dominatesfor all x ∼ < − , where it produces an enhancement, and specifically the enhancement of thegluon for x ∼ < .
03 which was discussed in Sect. 3.3. At large x instead the dominant effect isfrom the new data, which lead to a reduction of the gluon and an enhancement of the quarks.Whereas of course charm is very significantly affected by the change in methodology — it wasnot independently parametrized in NNPDF3.0 — for x ∼ > .
1, the new data also have a bigimpact. In fact, while strangeness is mostly affected by the new data in the medium and small x regions, charm and gluon are most affected by them at large x . Z boson The use of transverse momentum distributions has been advocated for a long time (see e.g.Ref. [2]) as a clean and powerful constraint on PDFs, particularly the gluon. As discussed inSection 2, it is now possible to include such data at NNLO thanks to the availability of thecomputation of this process up to NNLO QCD, along with precise data on
Z p T from ATLASand CMS at 8 TeV. The impact of this dataset on PDFS has recently been studied in detail inRef. [121].NNPDF3.1 is the first global PDF determination to include this data. In order to assess theimpact of this dataset, we have repeated the NNLO determination, excluding all Z p T data. InFig. 4.3 we show the distances between this PDF set and the default: it is clear that the effect onall PDFs is moderate, with changes below one third of a sigma. The largest differences are seenin the gluon, as expected, and the strange distributions. The reason for this state of affairs can bebest understood by directly comparing PDFs and their uncertainties, see Fig. 4.4. It is clear thatcentral values move very little while uncertainties are slightly reduced, therefore demonstratingthe excellent consistency of the constraint from these measurements with the existing dataset.The Z p T dataset therefore reinforces the reliability of our gluon determination. It also reducessomewhat the uncertainty on the total strangeness. While in Ref. [121] this dataset was found tohave a rather stronger impact than shown here, it should be noted that this was the case whendetermining PDFs from the NNPDF3.0 dataset (less the jet data). In NNPDF3.1 more dataare added, specifically top pair differential distributions: a smaller impact of the Z p T datasetwhen added to a wider prior is not unexpected.In addition to the 8 TeV measurements from ATLAS and CMS there also exists a mea-surement of the normalized distribution at 7 TeV from ATLAS. The inclusion of this dataset isproblematic because the covariance matrix for a normalized distribution depends on the cuts im-posed on the dataset, and only the covariance matrix for the full dataset is available. This issuewas studied in detail in Ref. [121]. Furthermore, this dataset is superseded by the more precise8 TeV measurement. Therefore, it has not been included in the NNPDF3.1 dataset. However,we have studied its potential impact by including it in a dedicated PDF determination, with itsnominal published covariance matrix unmodified despite the cuts.In Table 4.1 we provide the χ /N dat values for all the LHC Z p T measurements, for both theNNPDF3.1 NNLO baseline (including the ATLAS and CMS Z p T Z p T χ /N dat for the 7 TeV data is in parenthesis to indicate that, unlike all other values, itis a prediction and not the outcome of a fit. It is clear that the ATLAS Z p T W / Z rapidity distributions,whose total χ deteriorates by 12 units (with 46 datapoints). The distances between these twoPDF sets, displayed in Fig. 4.5, show that the gluon and quarks are shifted by almost one sigma42y the inclusion of the ATLAS Z p T Z p T dataset would have a significant impact on PDFs, without an improvement in precision,and with signs of tension between this dataset and both the remaining Z p T datasets, and other W and Z production data. Therefore its inclusion in the global dataset does not appear to bejustified. The impact of differential top pair production on PDFs and the optimal selection of top datasetshas been discussed extensively in Ref. [124]. Here we briefly study the impact of the top dataon NNPDF3.1 by comparing with PDFs determined removing the top data from the dataset.In Fig. 4.7 we show the distances between these PDF sets. Large differences can be seen in thegluon central value and uncertainty for x ∼ > .
1: these data constrain the gluon for values aslarge as x (cid:39) . x .The differences between the two PDF sets are demonstrated in Fig. 4.8, where the gluon andthe charm quark are shown. There is a substantial reduction in the uncertainty of the large x gluon, with the central value without top data being considerably higher than the narrow errorband of the result when top is included. This suggests a significant increase in the precisionof the gluon determination due to the top data. For the large x gluon the differences betweenNNPDF3.1 and NNPDF3.0 seen in Figs. 3.3-4.2 are therefore partly driven by the top data.The impact on quark PDFs is marginal, as can be seen in the case of charm.As already mentioned in Sect. 2.7, it has been shown in Ref. [125] that the sensitivity ofthe rapidity distribution on the top mass is minimal. In fact, in Ref. [124] it was shown thatif the top mass is varied by 1 GeV, NLO theoretical predictions for the normalized rapiditydistributions at the LHC 8 TeV vary by 0.6% at most in the kinematic range covered by thedata, which is much less than the uncertainty on the data, or the size of the NNLO corrections.This strongly suggests that our results are essentially independent of the value of the top mass. While jet data have been used for PDF determination for a long time, their full NNLO treatmentis only becoming possible now, thanks to the recent completion of the relevant computation [20,108]. However, as discussed in Section 2.4, NNLO corrections are not yet available for all datasetsincluded in NNPDF3.1. Consequently in the default NNPDF3.1 PDF determination, jets havebeen included using NNLO PDF evolution and NLO matrix elements supplemented by an extratheory uncertainty determined through scale variation. Here we assess generally the effect of jetdata, and in particular the possible impact of this approximation.
NNPDF3.1 NNLO + ATLAS
Z p T Z p T p llT , y ll ) [6.78] 3.40ATLAS Z p T p llT , M ll ) 0.93 0.98ATLAS Z p T p llT , y ll ) 0.93 1.17CMS Z p T p llT , M ll ) 1.32 1.33 Table 4.1:
The values of χ /N dat for the LHC Z p T data using the NNPDF3.1 NNLO PDF set, and fora new PDF determination which also includes the ATLAS Z p T
43o this end, we first repeat the NNPDF3.1 determination but excluding jet data. Thedistances between these PDFs and the default are shown in Fig. 4.9. It is clear that jet datahave a moderate and very localized impact, on the gluon in the region 0 . ∼ < x ∼ < .
6, at mostat the half-sigma level, with essentially no impact on other PDFs. The changes in all otherPDFs are compatible with a statistical fluctuation. A direct comparison of the gluon PDFs andtheir uncertainties in Fig. 4.10 confirms this. The uncertainty on the gluon is reduced by upto a factor of two by the jet data in this region, with the central value of the gluon within thenarrower uncertainty band of the default set.It is interesting to observe that in NNPDF3.0 the impact of the jet data was rather moresignificant, with uncertainties being reduced by a large factor for all x ∼ > .
1. In NNPDF3.1 thegluon at large x is strongly constrained by the top data, as discussed in Section 4.3. Specifi-cally, the addition of the jet data leaves the gluon unchanged in this region, see Fig. 4.10, butaddition of the top data produces a significant shift, as seen in Fig. 4.8. This suggests excellentcompatibility between the jet and top data, with the large x gluon now mostly determined bythe top data. This also explains the insensitivity to the NNLO correction to jet production, tobe discussed shortly.Despite there reduced impact, the jet data still play a non-negligible role. One may thereforeworry about the reliability of the theoretical treatment, based on NLO matrix elements withtheory uncertainties. In order to assess this, we have repeated the PDF determination but nowusing full NNLO theory in the case of the 2011 7 TeV LHC jet data where it is available. Theother jet datasets, namely the CDF Run II k T jets, the ATLAS and CMS √ s = 2 .
76 TeVdatasets, and the ATLAS 2010 7 TeV dataset, are treated as in the baseline. Essentially nochange in PDFs is found, as illustrated in Fig. 4.11 where the gluon and down PDFs are shown.Such a result is consistent with the percent level NNLO corrections found when using our choiceof the jet p T as the central scale, shown in Fig. 2.4.Also, as mentioned in Sect. 2.4, only the central rapidity bin of the ATLAS 2011 7 TeVdata has been included, because we have found that, while a good description can be achievedif each of the rapidity bins is included in turn, or if the uncertainties are decorrelated betweenrapidity bins, it is impossible to achieve a good description of all rapidity bins with correlationsincluded. One may therefore wonder whether the inclusion of other rapidity bins would leadto different results for the PDFs, despite the fact that they have less PDF sensitivity [104].In order to check this, we have compared to the data the prediction for all of the ATLAS2011 7 TeV data using the default NNPDF3.1 set, and determined the χ for each rapidity binseparately. For the five rapidity bins which have not been included, from central to forward,we find χ /N dat = 1 . , . , . , . , .
73, with respectively N dat = 29 , , , ,
12, tobe compared to the value χ /N dat = 1 .
06 for N dat = 31 of Tab 4.2 for the central rapidity binwhich is included. We conclude that all rapidity bins are well reproduced, and thus none ofthem can have a significant pull on PDFs that might change the result if they were included.In order to understand better the impact of the NNLO corrections and their effect on thePDFs, in Fig. 4.12 we compare the best-fit prediction to the 7 TeV 2011 CMS and ATLASdata for the two PDF sets compared in Figs. 4.9-4.11. The corresponding values of χ /N dat are collected in Table 4.2, both for these and all other jet data. First of all, one should notethat for all experiments for which an exact NNLO computation is not yet available, listed onthe top part of Table 4.2, the χ values obtained with these two PDF sets are almost identical.Because the predictions for these experiments are computed using the same theory (NLO withextra scale uncertainty), this shows that the change in PDFs is very small. For the experimentsfor which NNLO theory is available, χ values also change very little. A comparison of the datato theory (see Fig. 4.12) shows that the NLO and NNLO predictions are, as expected, quiteclose. Nevertheless, the NNLO prediction is in slightly better agreement with the data, and thisis reflected by the better value of the NNLO χ shown in Table 4.2, despite the the extra scale44ncertainty (shown as an inner error bar on the data in Fig. 4.12) that is added to the NLOprediction only.As a final check, we have repeated the PDF determination but with a cut excluding large p T jet data, for which the NNLO corrections are larger (compare Fig. 2.3). Specifically, we haveonly kept 7 TeV jet data with p T ≤
240 GeV, 2.76 TeV jet data with p T ≤
95 GeV, 1.96 TeV jetdata with p T ≤
68 GeV. We found that the ensuing PDFs are statistically equivalent (distancesof order one-two) to those from the default set. We conclude that the impact of the approximatetreatment of jet data on the default NNPDF3.1 set is very small.NNPDF3.1 exact NNLOCDF Run II k t jets 0.84 0.85ATLAS jets 2.76 TeV 1.05 1.03CMS jets 2.76 TeV 1.04 1.02ATLAS jets 2010 7 TeV 0.96 0.95ATLAS jets 2011 7 TeV 1.06 0.91CMS jets 7 TeV 2011 7 TeV 0.84 0.79Table 4.2: The values of χ /N dat for all the jet datasets obtained using either of the two PDF setscompared in Fig. 4.11, and in each case the theory used in the corresponding PDF determination. Forthe datasets in the top part of the table the exact NNLO computations are not yet available and NLOtheory with scale uncertainty is used throughout, while for those in the bottom part of the table NNLOtheory is used for the right column. x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q d ) / ( x , Q d NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q + ) / s ( x , Q + s NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q + ) / c ( x , Q + c NNLO, Q = 100 GeV
Figure 4.2:
Same as Fig. 3.3, but now also including PDFs determined using NNPDF3.1 methodologywith the NNPDF3.0 dataset. From left to right and from top to bottom the gluon, up, down, antidown,total strangeness and charm are shown. − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of
Z p T data Figure 4.3:
Same as Fig. 3.2, but now comparing the default NNPDF3.1 to a version of it with the 8TeV
Z p T data from ATLAS and CMS not included. x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNPDF3.1NNPDF3.1 no Z pT
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q + ) / s ( x , Q + s NNPDF3.1NNPDF3.1 no Z pT
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( g ( x , Q g ( x , Q d NNPDF3.1NNPDF3.1 no Z pT
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q + ) / ( s ( x , Q + s d NNPDF3.1NNPDF3.1 no Z pT
NNPDF3.1 NNLO, Q = 100 GeV
Figure 4.4:
Same as Fig. 3.3 (top) and as Fig. 3.4 (bottom), but now comparing the default NNPDF3.1to a version of it with the 8 TeV
Z p T data from ATLAS and CMS not included. Results are shown forthe gluon (left) and total strangeness (right). − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of ATLAS 7 TeV
Z p T data Figure 4.5:
Same as Fig. 3.2, but now comparing the default NNPDF3.1 to a version of it with the7 TeV
Z p T ATLAS data also included. x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNPDF3.1NNPDF3.1 + ATLAS Z pT 7 TeV
NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q NNPDF3.1NNPDF3.1 + ATLAS Z pT 7 TeV
NNLO, Q = 100 GeV
Figure 4.6:
Same as Fig. 3.3 but now comparing the default NNPDF3.1 to a version of it with the 7 TeV
Z p T ATLAS data also included. Results are shown for the gluon (left) and down quark (right). − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of top quark pair data
Figure 4.7:
Same as Fig. 4.3 but now excluding all top data (total cross-sections and differential distri-butions). Note the different scale on the y axis in the left plot. x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNPDF3.1NNPDF3.1 no top
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q + ) / c ( x , Q + c NNPDF3.1NNPDF3.1 no top
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( g ( x , Q g ( x , Q d NNPDF3.1NNPDF3.1 no top
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q + ) / ( c ( x , Q + c d NNPDF3.1NNPDF3.1 no top
NNPDF3.1 NNLO, Q = 100 GeV
Figure 4.8:
Same as Fig. 4.4 but now excluding all top data (total cross-sections and differential distribu-tions). Results are shown for the gluon (left) and charm (right), the PDFs above and their uncertaintiesbelow. − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of jet data
Figure 4.9:
Same as Fig. 4.3 but now excluding all jet data. x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNPDF3.1NNPDF3.1 no jet
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( g ( x , Q g ( x , Q d NNPDF3.1NNPDF3.1 no jet
NNPDF3.1 NNLO, Q = 100 GeV
Figure 4.10:
Comparison between the default NNPDF3.1 NNLO PDFs an alternative determination inwhich all jet data have been removed: the gluon (left) and the percentage uncertainty on it (right) areshown x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNPDF3.1NNPDF3.1 (NNLO jets)
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q NNPDF3.1NNPDF3.1 (NNLO jets)
NNPDF3.1 NNLO, Q = 100 GeV
Figure 4.11:
Same as Fig. 3.3 but now comparing the default NNPDF3.1 NNLO PDFs to an alternativedetermination in which ATLAS and CMS 7 TeV jet data have been included using exact NNLO theory.The gluon (left) and down (right) PDFs are shown.
50 500 750 1000 1250 1500 1750 2000p T (GeV)0.20.40.60.81.01.21.41.61.8 R a t i o t o D a t a CMS jets 7 TeV 2011 |y| = 0.25
DataNNPDF 3.1NNPDF 3.1 (NNLO Jets)
250 500 750 1000 1250 1500 1750p T (GeV)0.60.81.01.21.4 R a t i o t o D a t a ATLAS jets 2011 7 TeV |y| = 0.25
DataNNPDF 3.1NNPDF 3.1 (NNLO Jets)
Figure 4.12:
Comparison between CMS (left) and ATLAS (right) one-jet inclusive data at 7 TeV from2011, and best-fit results obtained using NLO theory supplemented by scale uncertainties or exact NNLOtheory. The uncertainties shown on the best-fit prediction is the PDF uncertainty, while that on the datais the diagonal (outer error bar) and the scale uncertainty on the NLO prediction (inner error bar). − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of LHCb data
Figure 4.13:
Same as Fig. 4.3 but now excluding all LHCb data. Note the different scale on the y axisin the left plot. Electroweak production data from the LHCb experiment open up a new kinematic region andtherefore provide new constraints on flavor separation at large and small x . While LHCb datawere included in NNPDF3.0, the release of the legacy Run I LHCb measurements at 7 TeV and8 TeV, which include all correlations between W and Z data, greatly increases their utility inPDF determination. In Fig. 4.13 distances are shown between the NNPDF3.1 NNLO defaultand PDFs determined excluding all LHCb data. The impact is significant for all quark PDFs,especially in the valence region: hence this data has a substantial impact on flavor separation,most notably at large x .This is explicitly demonstrated for the up, down and charm PDFs in Fig. 4.14, where thepercentage PDF uncertainties are also shown. The LHCb data play a significant role in thedata-driven large- x enhancement of light quark PDFs discussed in Section 4.1 (see Fig. 4.2),and is largely responsible for the sizable impact of new data on charm for x ∼ > .
1, where theysignificantly reduce the uncertainty. Effects are more marked at medium and large x , peaking at51 x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q NNPDF3.1NNPDF3.1, no LHCb
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) ) / u ( x , Q u ( x , Q d NNPDF3.1NNPDF3.1, no LHCb
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q NNPDF3.1NNPDF3.1, no LHCb
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) ) / d ( x , Q d ( x , Q d NNPDF3.1NNPDF3.1, no LHCb
NNPDF3.1 NNLO, Q = 100 GeV x - - ) [ r e f] ( x , Q + ) / c ( x , Q + c NNPDF3.1 NNLO, Q = 100 GeV x - - ) ( x , Q + ) / c ( x , Q + c d NNPDF3.1 NNLO, Q = 100 GeV
Figure 4.14:
Same as Fig. 4.4 but now excluding all LHCb data. Results are presented, from top tobottow, for the up, down and charm PDFs. Both PDFs (left) and uncertainties (right) are shown. around x (cid:39) .
3: in this region the PDF uncertainty is also substantially reduced; the reductionin uncertainty is especially marked for the down PDF.In order to see the impact of the LHCb data directly, in Fig. 4.15 we compare the 8 TeVLHCb muon W + and W − data to predictions obtained using NNPDF3.0 and NNPDF3.1. Theimprovement is clear, particularly for large rapidities. There is also a noticeable reduction inPDF uncertainty on the prediction. W asymmetries from the Tevatron W production data from the Tevatron have for many years been the leading source of informationon quark flavor decomposition. The final legacy D0 W asymmetry measurements in the electronand muon channels are included in NNPDF3.1, superseding all previous data. In Fig. 4.16 weperform a distance comparison between the default NNPDF3.1 and PDFs determined excludingthis dataset. Distances are generally small, an observation confirmed by direct PDF comparisonin Fig. 4.17. However, we have seen in Tab. 3.1 that the fit quality for this dataset is ratherbetter with NNPDF3.1 than with the previous NNPDF3.0. The moderate impact of this dataset52 .50 2.75 3.00 3.25 3.50 3.75 4.00 4.250.8500.8750.9000.9250.9500.9751.0001.025 R a t i o t o D a t a LHCb W, Z 8 TeV (W + ) DataNNPDF 3.1NNPDF 3.0 R a t i o t o D a t a LHCb W, Z 8 TeV (W )
DataNNPDF 3.1NNPDF 3.0
Figure 4.15:
Comparison between 8 TeV LHCb muon W + (left) and W − (right) production data toNNLO predictions obtained using NNPDF3.1 and NNPDF3.0. The uncertainties shown are the diagonalexperimental uncertainty for the data, and the PDF uncertainty for the best-fit prediction. − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of D0 W asymmetry data Figure 4.16:
Same as Fig. 4.3 but now excluding D0 W asymmetry data. is due to its excellent consistency with the abundant LHC data, which are now driving flavorseparation. This data thus provides further evidence for the reliability of the flavor separationin NNPDF3.1. W, Z production data and strangeness
ATLAS W and Z production data were already included in NNPDF3.0, but recent measurementsbased on the 2011 dataset [72] have much smaller statistical uncertainties. This dataset, like theprevious ATLAS measurement, has been claimed to have a large impact on strangeness. This isborne out by the plot, Fig. 4.18, of the distance between the default NNPDF3.1 and a versionfrom which this dataset has been excluded. Indeed the largest effect — almost at the one sigmalevel on central values — is seen on the strange and charm PDFs, with a rather smaller impacton all other PDFs.A direct comparison of the strange and charm PDFs in Fig. 4.19 shows that strangeness issignificantly enhanced in the medium/small x region by the inclusion of the ATLAS data, whilecharm is suppressed. As discussed in Section 3.4 and shown in Fig. 3.8, this suppression of charmcannot be accommodated when charm is perturbatively generated, and therefore parametrizing53 x - - - - ) [ r e f] ( x , Q u ) / ( x , Q u NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q d ) / ( x , Q d NNPDF3.1 NNLO, Q = 100 GeV
Figure 4.17:
Same as Fig. 4.4 but now excluding D0 W asymmetries. The antiup (left) and antidown(right) PDFs are shown. − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of
W, Z
Figure 4.18:
Same as Fig. 4.3 but now excluding 2011 ATLAS
W, Z rapidity distributions. charm is important in order to be able to reconcile the ATLAS data with the global datasetwhich in this x range is severely constrained by HERA data. The strange and charm content ofthe proton will be discussed in detail in Sections 5.2 and 5.3 below.NNPDF3.1 achieves a good description of the data, as illustrated in Fig. 4.20 where theNNLO prediction obtained using NNPDF3.0 is also shown. It is clear that the agreement isgreatly improved, as demonstrated in the χ values shown in Tab. 3.1. It is also interestingto note the significant reduction in PDF uncertainties. As mentioned in Sect. 2.5, these datahave been included only partially in the NNPDF3.1 determination: specifically Z productiondata off peak or at forward rapidity have not been included. We have checked however that thedescription of these data in NNPDF3.1 is equally good, and similarly improved in comparisonto NNPDF3.0, with χ /N dat of order unity. A comparison of these data for two of the four binswhich have not been included to the NNPDF3.1 and NNPDF3.0 is shown in Fig. 4.21. Like ATLAS, CMS has also published updated electroweak boson production data. The NNPDF3.0PDF determination already included double-differential (in rapidity and invariant mass) Drell-54 x - - - - ) [ r e f] ( x , Q + ) / s ( x , Q + s NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q + ) / c ( x , Q + c NNLO, Q = 100 GeV
Figure 4.19:
Same as Fig. 4.4 but now excluding 2011 ATLAS
W, Z rapidity distributions. The totalstrange (left) and charm (right) PDFs are shown. R a t i o t o D a t a ATLAS W, Z 7 TeV 2011 (W )
DataNNPDF 3.1NNPDF 3.0 R a t i o t o D a t a ATLAS W, Z 7 TeV 2011 (Z)
DataNNPDF 3.1NNPDF 3.0
Figure 4.20:
Comparison between the 2011 ATLAS 7 TeV W − (left) and Z (right) data to NNLOpredictions obtained using NNPDF3.1 and NNPDF3.0; W production data are plotted versus the pseu-dorapidity of the forward lepton η l , while Z production data are plotted vs the dilepton rapidity y ll . Yan data at 7 TeV from the CMS 2011 dataset [54]. An updated version of the same mea-surement at 8 TeV based on 2012 data was presented in Ref. [84], including both the absolutecross-sections and the ratio of 8 TeV and 7 TeV measurements.This data has very small uncorrelated systematic uncertainties. Unfortunately, only thefull covariance matrix, with no breakdown of individual correlated systematics, has been madeavailable. The combination of these two facts makes it impossible to include this experimentin the NNPDF3.1 dataset, as we now explain. In Fig. 4.22 we show the distances between theNNPDF3.1 NNLO PDF set and a modified version of it where this dataset has been included.While the impact on uncertainties is moderate, clearly this dataset has a significant impact at thelevel of central-values on all PDFs for almost all x values, with a particularly important impacton the medium/small x gluon. This is somewhat surprising, given that Drell-Yan productiononly provides an indirect handle on the gluon PDF.A direct comparison of PDFs and their uncertainties in Fig. 4.23 shows that these datainduce an upwards shift by up to one sigma of the gluon for x ∼ < .
1, and a downward shift ofthe light quark PDFs for x ∼ > .
1, by a comparable amount. This, however, is not accompaniedby a reduction of PDF uncertainties, which increase a little, as also shown in Fig. 4.23.Furthermore, while the fit quality of the 8 TeV CMS double-differential Drell-Yan data re-mains poor after their inclusion in the fit, with a value of χ /N dat = 2 .
88, there is a certaindeterioration in fit quality of all other experiments. Indeed, the total χ to all the other data55 .25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25| /y|0.900.920.940.960.981.001.021.04 R a t i o t o D a t a ATLAS W, Z 7 TeV 2011 (Z high ) DataNNPDF 3.1NNPDF 3.0 R a t i o t o D a t a ATLAS W, Z 7 TeV 2011 (Z peak ) DataNNPDF 3.1NNPDF 3.0
Figure 4.21:
Same as Fig. 4.20 but now for two of the four data bins which have not been included in theNNPDF3.1 determination: high-mass Z production at central rapidity (left) and on-shell Z productionat forward rapidity (right). − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of CMS 8 TeV DY double differential data
Figure 4.22:
Same as Fig. 3.2 but now comparing the default NNPDF3.1 to a version of it with the8 TeV CMS double-differential Drell-Yan data also included. deteriorates by ∆ χ = 11 .
5. A more detailed inspection shows that the most marked deterio-ration is seen in the HERA combined inclusive DIS data, with ∆ χ = 19 .
7. This means thatthere is tension between the CMS data and the rest of the global dataset, and more specificallytension with the HERA data, which are most sensitive to the small x gluon.We must conclude that this experiment appears to be inconsistent with the global dataset,and particularly with the data with which we have the least reasons to doubt, namely thecombined HERA data and their determination of the gluon. In the absence of more detailedinformation on the covariance matrix it is not possible to further investigate the matter, andthe 8 TeV CMS double-differential Drell-Yan data have consequently not been included in theglobal dataset. F c data and intrinsic charm. The advantages of introducing an independently parametrized charm PDF were advocated inRef. [23], where a first global PDF determination including charm was presented, based on theNNPDF3.0 methodology and dataset. The default NNPDF3.0 dataset was supplemented bycharm deep-inelastic structure function data from the EMC collaboration [69]. This dataset is56 x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( g ( x , Q g ( x , Q d NNPDF3.1NNPDF3.1+CMSDY8TEV
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( u ( x , Q u ( x , Q d NNPDF3.1NNPDF3.1+CMSDY8TEV
NNPDF3.1 NNLO, Q = 100 GeV
Figure 4.23:
Same as Fig. 3.3 (top) but now comparing the default NNPDF3.1 to a version of it with the8 TeV CMS double-differential Drell-Yan data also included. The corresponding percentage uncertaintiesare also shown (bottom). Results are shown for the gluon (left) and up quark (right). quite old, but it remains the only measurement of the charm structure function in the large x region. With the wider NNPDF3.1 dataset the EMC dataset is no longer quite so indispensable,specifically in view of phenomenology at the LHC: it has thus been omitted from the defaultNNPDF3.1 determination as doubts have been raised about its reliability.However, a number of checks performed in Refs. [23, 95], such as variations of kinematicalcuts and systematic uncertainties, do not suggest any serious compatibility issues, and ratherconfirmed this dataset as being as reliable as the other older datasets with fixed nuclear targetsroutinely included in global PDF determinations. Therefore it is interesting to revisit the issue ofthe impact of this dataset within the context of NNPDF3.1. To this purpose we have produceda modified version of the global NNPDF3.1 NNLO analysis in which the EMC dataset [69]is added to the default NNPDF3.1 dataset. In Fig. 4.24 the distances between this PDF setand the default are shown. The EMC dataset has a non-negligible impact on charm, at theone sigma level, and also to a lesser extent, at the half-sigma level, on all light quarks, withonly the gluon left essentially unaffected. The bulk of the effect is localized in the region0 . ∼ < x ∼ < .
3. The PDFs are directly compared in Fig. 4.25. The EMC data lead to anincrease in the charm distribution towards the upper edge of its error band in the default PDFset for 0 . ∼ < x ∼ < .
2, while reducing the uncertainty on it by a sizable factor. The light quarkPDFs are correspondingly slightly suppressed, and their uncertainties also reduced a little.The values of χ /N dat for the deep-inelastic scattering experiments and for the total datasetbefore and after inclusion of the EMC data in the NNPDF3.1 dataset are shown in Table 4.3.The fit quality of the EMC charm dataset is greatly improved by its inclusion, without anysignificant change of the fit quality for any other DIS data: the hadron collider data are even57 − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of EMC charm data
Figure 4.24:
Same as Fig. 4.5, but now comparing the default NNPDF3.1 to a version of it with theEMC F c dataset also included. less sensitive. The inclusion of EMC charm data therefore appears to give a more accuratecharm determination, with no cost elsewhere, and so usage of this PDF set is recommendedwhen precise charm PDFs at large x are required. The phenomenological implications of thecharm PDF will be discussed in Section 5.3 below. NNPDF3.1 NNLO + EMC charm dataNMC 1.30 1.29SLAC 0.75 0.76BCDMS 1.21 1.24CHORUS 1.11 1.10NuTeV dimuon 0.82 0.88HERA I+II inclusive 1.16 1.16HERA σ NC c F b F c [4.8] 0.93 Total dataset 1.148 1.145
Table 4.3:
The values of χ /N dat for the deep-inelastic scattering experiments, as well as for the totaldataset, for the NNPDF3.1 NNLO PDF set and for a new PDF determination which also includes theEMC charm structure function data. x - - - - ) [ r e f] ( x , Q + ) / c ( x , Q + c NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q + ) / ( c ( x , Q + c d NNPDF3.1 NNPDF3.1 + EMC
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q NNPDF3.1 NNLO, Q = 100 GeV
Figure 4.25:
Same as Fig. 4.6 but now comparing the default NNPDF3.1 to a version of it with theEMC F c dataset also included. Results are shown for the charm (top left), up (bottom left) and down(bottom right) PDFs. The relative PDF uncertainty on charm is also shown (top right). We have seen that the LHC data have a significant impact on various PDFs. Both in order toprecisely gauge this impact, and in view of possible applications in which usage of PDFs withoutLHC data is required, we have produced a PDF set in which all LHC data are excluded fromthe NNPDF3.1 dataset. The distance between the ensuing PDF set and the default are shownin Fig. 4.26.The cumulative effect of the data which were discussed in Sects. 4.2—4.5 and 4.7 is consid-erable. Most PDFs are affected at the one-sigma level and in some cases (such as the down andcharm quarks) at up to the two-sigma level. This is confirmed by direct comparison of the PDFs,see Fig. 4.27. The difference between the two fits appears to be mostly driven by the CHORUS,BCDMS and fixed-target Drell-Yan data, whose χ improves respectively by 84, 32 and 38 unitswhen removing the LHC data. Other datsets display much more smaller differences, typicallycompatible with statistical fluctuations, and in some cases (such as for the SLAC data) the fitquality is actually somewhat better in the global fit.On the other hand, it is clear that the shifts between PDFs without LHC data and thoseincluding them are compatible with the respective PDF uncertainties, and that the uncertaintieson the PDFs determined without LHC data are not so large as to render them useless forphenomenology. We conclude that the default set remains considerably more accurate andshould be used for precision phenomenology. However, the use of PDFs determined withoutsome or all the LHC data may be mandatory in searches for new physics, in order to make surethat possible new physics effects are not reabsorbed in the PDFs. In such circumstances, weconclude that even though the uncertainty in the PDFs without LHC data is not competitive, the59 − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of LHC data
Figure 4.26:
Same as Fig. 4.3 but now excluding all LHC data. level of deterioration is not so great as to make searches for new physics altogether impossible.
The NNPDF3.1 dataset includes several measurements taken upon nuclear targets. DIS datafrom the SLAC, BCDMS and NMC experiments along with the E886 fixed-target Drell-Yan datainvolve measurements of deuterium. All neutrino data and the fixed-target E605 Drell-Yan data,are obtained with heavy nuclear targets. All of these data were already included in previousPDF determinations, including NNPDF3.0. The impact of nuclear corrections was studied inRef. [5] and found to be under control. However, the much wider dataset might now permit theremoval of these data from the global dataset: whereas removing data inevitably entails someloss of precision, this might be more than compensated by the increase in accuracy due to thecomplete elimination of any dependence on uncertain nuclear corrections.In order to assess this, we performed two additional PDF determinations with the NNPDF3.1methodology. Firstly, by removing all heavy nuclear target data but keeping deuterium data,and secondly removing all nuclear data and only keeping proton data. The distances betweenthe default and these two PDF sets are shown in Fig. 4.28. At large x the impact of nucleartarget data is significant, at the one to two sigma level, mostly on the flavor separation of thesea. The deuterium data also have a significant impact, particularly in the intermediate x range.A direct comparison of PDFs, in Fig. 4.29, and their uncertainties, in Fig. 4.32, shows thatindeed PDFs determined with no heavy nuclear target data are reasonably compatible with theglobal set, though with rather larger uncertainties, especially for strangeness. Indeed, best-fitresults without heavy nuclear targets, or even without deuterium data, are all compatible withintheir respective uncertainties, which is consistent with the previous conclusion that the absenceof nuclear corrections for these data does not lead to significant bias at the level of current PDFuncertainties. On the other hand, PDFs determined with only proton data while compatible towithin one sigma with the global set within their larger uncertainties, show a substantial lossof precision. This is particularly notable for down quarks, due to the importance of deuteriumdata in pinning down the isospin triplet PDF combinations.Because deuterium data have a significant impact on the fit, one may worry that nuclearcorrections to the deuterium data are now no longer negligible, at the accuracy of the presentPDF determination. In order to investigate this issue in greater detail, we have performed a60 x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q NNPDF3.1NNPDF3.1 no LHC
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q NNPDF3.1NNPDF3.1 no LHC
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q + ) / c ( x , Q + c NNPDF3.1NNPDF3.1 no LHC
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNPDF3.1NNPDF3.1 no LHC
NNPDF3.1 NNLO, Q = 100 GeV
Figure 4.27:
Same as Fig. 4.4 but now excluding all LHC data. Results are shown for the up (top left),down (top right), charm (bottom left) and gluon (bottom right) PDFs. variant of the NNPDF3.1 NNLO default PDF determination in which all deuterium data arecorrected using the same nuclear corrections as used by MMHT14 (specifically, Eqs. (9,10) ofRef. [7]).In terms of fit quality we find that the inclusion of nuclear corrections leads to a slightdeterioration in the quality of the fit, with a value of χ /N dat = 1 . χ /N dat = 1 .
148 (see Table 3.1). In particular we find that for the NMC, SLAC,and BCDMS data the values of χ /N dat with (without) nuclear corrections are respectively0.94(0.95), 0.71(0.70), and 1.11(1.11). Therefore, the addition of deuterium corrections has nosignificant impact on the fit quality to these data.The distances between PDFs determined including deuterium corrections and the default areshown in Fig. 4.30. They are seen to be moderate and always below the half-sigma level, andconfined mostly to the up and down PDFs, as expected. These PDFs are shown in Fig. 4.31,which confirms the moderate effect of the deuterium correction. It should be noticed that thePDF uncertainty, also shown in Fig. 4.31, is somewhat increased when the deuterium correctionsare included. The relative shift for other PDFs are yet smaller since they are affected by largeruncertainties, which are also somewhat increased by the inclusion of the nuclear corrections.In view of the theoretical uncertainty involved in estimating nuclear corrections, and bearingin mind that we see no evidence of an improvement in fit quality while we note a slight increasein PDF uncertainties when including deuterium corrections using the model of Ref. [7], weconclude that the impact of deuterium corrections on the NNPDF3.1 results is sufficiently smallthat they may be safely ignored even within the current high precision of PDF determination.Nevertheless, more detailed dedicated studies of nuclear corrections, also in relation to theconstruction of nuclear PDF sets, may well be worth pursuing in future studies.61 − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of nuclear fixed-target data − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of nuclear+deuteron fixed-target data
Figure 4.28:
Same as Fig. 4.3 but now excluding all data with heavy nuclear targets, but keepingdeuterium data (top) or excluding all data with any nuclear target and only keeping proton data (bottom)
In conclusion, for the time being it is still appears advantageous to retain nuclear target datain the global dataset for general-purpose PDF determination. However, if very high accuracy isrequired (such as, for instance, in the determination of standard model parameters) it might bepreferable to use PDF sets from which all data with nuclear targets have been omitted.
A yet more conservative option to that discussed in the previous Section is to retain only colliderdata from HERA, the Tevatron and the LHC. The motivation for this suggestion, first presentedin the NNPDF2.3 study [133], is that this excludes data taken at low scales, which may be subjectto potentially large perturbative and non-perturbative corrections. Furthermore, data taken onnuclear targets, and all of the older datasets are eliminated, thereby leading to a more reliableset of PDFs. However, previous collider-only PDFs had very large uncertainties, due to thelimited collider dataset then available.In order to re-assess the situation with the current, much wider LHC dataset, we haverepeated a collider-only PDF determination. This amounts to repeating the proton only PDF62 x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNPDF3.1 NNPDF3.1 no nuclearNNPDF3.1 protonNNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q NNPDF3.1 NNPDF3.1 no nuclearNNPDF3.1 protonNNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q NNPDF3.1 NNPDF3.1 no nuclearNNPDF3.1 protonNNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q d ) / ( x , Q d NNPDF3.1 NNPDF3.1 no nuclearNNPDF3.1 protonNNPDF3.1 NNLO, Q = 100 GeV
Figure 4.29:
Same as Fig. 4.4 but now excluding all data with heavy nuclear targets, but keepingdeuterium data, or excluding all data with any nuclear target and only keeping proton data. Results areshown for the gluon (top left), up (top right), down (bottom left) and antidown (bottom right). determination described in the previous Section, but now with the proton fixed target data alsoremoved. The distances between the ensuing PDFs and the default NNPDF3.1 are shown inFig. 4.33. Comparing to Fig. 4.28 we notice that distances are similar for most PDFs, the mainexception being the gluon, which in the intermediate x region has now shifted by almost twosigma.This is confirmed by a direct comparison of PDFs in Fig. 4.34 and their uncertainties inFig. 4.32. The valence quarks (especially up) are reasonably stable, but the sea now is quiteunstable upon removal of all the fixed target data, visibly more than in the proton only setFig. 4.29, with in particular a substantial increase in the uncertainty of the anti-up quarkdistribution at large x . Furthermore, the gluon, which in Fig. 4.29 was quite stable in the proton-only PDF set, now undergoes a significant downward shift at intermediate x , even though itsuncertainty is not substantially increased.We conclude that, despite impressive improvements due to recent LHC measurements, acollider-only PDF determination is still not very useful for general phenomenological applica-tions. 63 − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Impact of deuteron corrections
Figure 4.30:
Same as Fig. 4.3 but now comparing the default NNPDF3.1 to a version in which alldeuterium data have been corrected using the nuclear corrections from Ref. [7]. x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q Baselinewith deut corr
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q Baselinewith deut corr
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( u ( x , Q u ( x , Q d Baselinewith deut corr
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( d ( x , Q d ( x , Q d Baselinewith deut corr
NNPDF3.1 NNLO, Q = 100 GeV
Figure 4.31:
Same as Fig. 4.4 but now comparing the default NNPDF3.1 to a version in which alldeuterium data have been corrected using the nuclear corrections from Ref. [7]. Results are shown forthe up (left= and down (right) PDFs. The uncertainties are also shown (botton row). x - - - - ) [ r e f] ) ) / ( u ( x , Q u ( x , Q d NNPDF3.1 NNPDF3.1 no nuclearNNPDF3.1 protonNNPDF3.1 collider
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q u ) / ( ( x , Q u d NNPDF3.1 NNPDF3.1 no nuclearNNPDF3.1 protonNNPDF3.1 collider
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( d ( x , Q d ( x , Q d NNPDF3.1 NNPDF3.1 no nuclearNNPDF3.1 protonNNPDF3.1 collider
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q d ) / ( ( x , Q d d NNPDF3.1 NNPDF3.1 no nuclearNNPDF3.1 protonNNPDF3.1 collider
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( s ( x , Q s ( x , Q d NNPDF3.1 NNPDF3.1 no nuclearNNPDF3.1 protonNNPDF3.1 collider
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q s ) / ( ( x , Q s d NNPDF3.1 NNPDF3.1 no nuclearNNPDF3.1 protonNNPDF3.1 collider
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ( x , Q + ) / ( c ( x , Q + c d NNPDF3.1 NNPDF3.1 no nuclearNNPDF3.1 protonNNPDF3.1 collider
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) ) / ( g ( x , Q g ( x , Q d NNPDF3.1 NNPDF3.1 no nuclearNNPDF3.1 protonNNPDF3.1 collider
NNPDF3.1 NNLO, Q = 100 GeV
Figure 4.32:
Comparison of the relative PDF uncertainties at Q = 100 between the NNPDF3.1 andthe no heavy nuclei, proton–only and collider–only PDF determinations. The uncertainties shown are allnormalized to the NNPDF3.1 central value. − − − − x d [ x , Q ] Central Value gdu ¯ d ¯ us + c + − − − − x Q = 100 GeV Uncertainty
NNPDF3.1 NNLO, Global vs collider-only fit
Figure 4.33:
Same as Fig. 4.3 but now only keeping collider data. x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q NNPDF3.1NNPDF3.1 collider-only
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q NNPDF3.1NNPDF3.1 collider-only
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q NNPDF3.1NNPDF3.1 collider-only
NNPDF3.1 NNLO, Q = 100 GeV x - - - - ) [ r e f] ( x , Q d ) / ( x , Q d NNPDF3.1NNPDF3.1 collider-only
NNPDF3.1 NNLO, Q = 100 GeV
Figure 4.34:
Same as Fig. 4.4 but now only keeping collider data. Results are shown for the gluon (topleft), up (top right), down (bottom left) and antidown (bottom right). Implications for phenomenology
We now present some initial studies of the phenomenological implications of NNPDF3.1 PDFs.Firstly, we summarize the status of PDF uncertainties building upon the discussion in theprevious Section; we discuss the status of PDF uncertainties, and then we focus on the strangeand charm PDF. We then discuss PDF luminosities which are the primary input to hadroncollider processes, and predictions for LHC processes, specifically W , Z and Higgs productionat the LHC. As elsewhere, only a selection of results is presented here, with a much larger set isavailable online as discussed in Section 6.2. After discussing in Section 4 the impact on NNPDF3.1 PDFs of each individual new piece ofdata and after separating off the effect of the new methodology, we can now study the combinedeffect of all the new data by comparing PDF uncertainties in NNPDF3.0 and NNPDF3.1. Thisis done in Fig. 5.1, where we compare relative PDF uncertainties (all computed with a commonnormalization) on PDFs in the NNPDF3.0 and NNPDF3.1 sets, shown as valence (i.e. q − ¯ q ),and sea (i.e. ¯ q ) for up and down and q + ¯ q for strange and charm. Results are shown both forindividual PDF flavors, and for the singlet, valence, and triplet combinations defined in Eq. (3.4)of Ref. [23]).The most visible effect is the very considerable reduction in gluon uncertainty, which is nowat the percent level for almost all x . As discussed in Section 4, this is due to the combination ofmany mutually consistent constraints on the gluon from DIS (especially at HERA), Z transversemomentum distributions, jet production, and top pair production, which taken together cover avery wide kinematic range. The singlet quark combination, which mixes with the gluon, showsa comparable improvement for all x ∼ < .
1, but less marked at large x .Interestingly, for quark PDFs the pattern of uncertainties is different in the flavor basisversus the “evolution” basis as given in Eq. (3.4) of Ref. [23]). Specifically, the aforementionedreduction in uncertainty on the singlet combination is not seen in any of the light quark valencedistributions, which generally have comparable uncertainties in NNPDF3.1 and NNPDF3.0.This is due to the fact that flavor separation is somewhat more uncertain in NNPDF3.1, due tothe fact that charm is now independently parametrized. This is compensated by the availabilityof more experimental information (in particular LHCb and ATLAS data), but not at smalland very large x . Indeed, the valence and triplet distributions have generally somewhat largeruncertainties in NNPDF3.1 than in NNPDF3.0, except for x ∼ . x ∼ < − they are all comparable, andall (except for strangeness) rather smaller than the corresponding uncertainties in NNPDF3.0,since they are driven by the mixing of the singlet and the gluon through perturbative evolutionat small x [134]. Even the charm PDF, now independently parametrized, has a smaller uncer-tainty. In the large- x region, instead, the less accurate knowledge of flavor separation kicks in,and relative uncertainties are larger. Clearly, in NNPDF3.0 charm had an unnaturally smalluncertainty, since at large x perturbatively-generated charm is tied to the gluon. In NNPDF3.1,instead, the hierarchy of uncertainties on sea PDFs in the valence region is what one wouldexpect, with up and down known most accurately, and strange and charm affected by increas-ingly large uncertainties. The uncertainty in NNPDF3.1 on the up and especially down seacomponents is a little increased, but still comparable to NNPDF3.0, while the uncertainty instrangeness is stable and that on charm significantly increased, as it should be given that large- x charm is largely unconstrained by data. In this respect, it is interesting to observe that a moreaccurate determination of charm and other sea PDFs at large x can be achieved through theinclusion of the EMC dataset as discussed in Section 4.9 above. Future LHC data on processes67 x - - - - R e l a t i v e P D F un c e r t a i n t i e s NNPDF3.0 NNLO, Q = 100 GeVUp ValenceDown ValenceGluonNNPDF3.0 NNLO, Q = 100 GeV x - - - - R e l a t i v e P D F un c e r t a i n t i e s NNPDF3.1 NNLO, Q = 100 GeVUp ValenceDown ValenceGluonNNPDF3.1 NNLO, Q = 100 GeV x - - - - R e l a t i v e P D F un c e r t a i n t i e s NNPDF3.0 NNLO, Q = 100 GeV
Up SeaDown SeaStrange SeaCharm Sea
NNPDF3.0 NNLO, Q = 100 GeV x - - - - R e l a t i v e P D F un c e r t a i n t i e s NNPDF3.1 NNLO, Q = 100 GeV
Up SeaDown SeaStrange SeaCharm Sea
NNPDF3.1 NNLO, Q = 100 GeV x - - - - R e l a t i v e P D F un c e r t a i n t i e s NNPDF3.0 NNLO, Q = 100 GeV S TV NNPDF3.0 NNLO, Q = 100 GeV x - - - - R e l a t i v e P D F un c e r t a i n t i e s NNPDF3.1 NNLO, Q = 100 GeV S TV NNPDF3.1 NNLO, Q = 100 GeV
Figure 5.1:
Comparison of relative uncertainties on NNPDF3.0 (left) and NNPDF3.1 (right) NNLOPDFs, normalized to the NNPDF3.1 NNLO central value. The two light quark valence PDFs and thegluon are shown (top) along with all individual sea PDFs (center) and the singlet, valence and isospintriplet combinations (bottom). such as Z + c production may confirm the reliability of the EMC dataset. Whereas there is broad consensus on the size, i.e. the central value, of up and down PDFs, forwhich there is good agreement between existing determinations within their small uncertainty,the size of the strange PDF has been the object of some controversy, which we revisit here inview of NNPDF3.1 results. Specifically, the strange fraction of proton quark sea, defined as R s ( x, Q ) = s ( x, Q ) + ¯ s ( x, Q )¯ u ( x, Q ) + ¯ d ( x, Q ) . (5.1)68nd the corresponding ratio of momentum fractions K s ( Q ) = (cid:82) dx x (cid:0) s ( x, Q ) + ¯ s ( x, Q ) (cid:1)(cid:82) dx x (cid:0) ¯ u ( x, Q ) + ¯ d ( x, Q ) (cid:1) , (5.2)have been traditionally assumed to be significantly smaller than one, and in PDF sets producedbefore the strange PDF could be extracted from the data, such as e.g. NNPDF1.0 [97], it wasoften assumed that R s ∼ , for all x , and thus also K s ∼ . This level of strangeness suppressionis indeed found in many recent global PDF sets, in which the strongest handle on the strangePDF is provided by deep-inelastic neutrino inclusive F and charm F c (“dimuon”) data.This was challenged in Ref. [135] where, on the basis of ATLAS W and Z production data,combined with HERA DIS data, it was argued instead that, in the measured region, the strangefraction R s is of order one. In Refs. [3,133], respectively based on the NNPDF2.3, and NNPDF3.0global analyses, both of which includes the data of Ref. [135], it was concluded that whereas theATLAS data do favor a larger strange PDF, they have a moderate impact on the global PDFdetermination due to large uncertainties, and also, that if the strange PDF is only determinedfrom HERA and ATLAS data, the central value is consistent with the conclusion of Ref. [135],but the uncertainty is large enough to lead to agreement with the suppressed strangeness ofthe global PDF sets to within one sigma. In Ref. [3] it was also shown that the CMS W + c production data [60], which were included there for the first time and which are also included inNNPDF3.1, though only in the NLO determination because of lack of knowledge of the NNLOcorrections, have a negligible impact due to their large uncertainties.As we discussed in Section 4.7, ATLAS W and Z production data have been supplementedby the rather more accurate dataset of Ref. [72], also claimed to favor enhanced strangeness.Indeed, we have seen in Section 4.7 that strangeness is significantly enhanced by the inclusionof these data, and also, in Section 3.4, that this enhancement can be accommodated in theglobal PDF determination thanks to the independently parametrized charm PDF, which is anew feature to NNPDF3.1. It is thus interesting to re-asses strangeness in NNPDF3.1, bycomparing theoretically motivated choices of dataset: we will thus compare to the previousNNPDF3.0 results for strangeness obtained using the default NNPDF3.1, the collider-only PDFset of Section 4.12, which can be considered to be theoretically more reliable, and a PDF setwhich we have constructed by using NNPDF3.1 methodology, but only including all HERAinclusive structure function data from Tab. 2.1 and the ATLAS data of Ref. [72]. Becauseinclusive DIS data alone cannot determine separately strangeness [1] this is then a determinationof strangeness which fully relies on the ATLAS data.In Table 5.1 we show NNLO results, obtained using these different PDF sets, for R s ( x, Q )Eq. 5.1 at Q = 1 .
38 GeV (thus below charm threshold) and Q = m Z and x = 0 . x value chosen by ATLAS in order to maximize sensitivity. Results are also compared to that ofRef. [72]. A graphical representation of the table is in Fig. 5.2.First, comparison of the NNPDF3.1 HERA+ATLAS W, Z result with that of Ref. [72], basedon the same data, shows agreement at the one-sigma level, with a similar central value and agreatly increased uncertainty, about four times larger, most likely because of the more flexibleparametrization and because of independently parametrizing charm. Second, strangeness inNNPDF3.1 is rather larger than in NNPDF3.0: as we have shown in Sects. 3.4,4.7 this is largelydue to the effect of the ATLAS
W, Z
DF set R s (0 . , .
38 GeV) R s (0 . , M Z )NNPDF3.0 0.45 ± ± ± ± ± ± W, Z ± ± xFitter HERA + ATLAS
W, Z (Ref. [72]) 1 . +0 . − . - Table 5.1:
The strangeness fraction R s ( x, Q ) Eq. (5.1) at x = 0 . W, Z sets, compared to the xFitter
ATLAS value Ref. [72]. ) = 1.9 GeV (x=0.023, Q S R )d + u ) / ( s = ( s + S R NNPDF3.0NNPDF3.1 globalNNPDF3.1 colliderNNPDF3.1 HERA+AWZ11xFitter 2016 )d + u ) / ( s = ( s + S R ) = M (x=0.023, Q S R )d + u ) / ( s = ( s + S R NNPDF3.0NNPDF3.1 globalNNPDF3.1 colliderNNPDF3.1 HERA+AWZ11 )d + u ) / ( s = ( s + S R Figure 5.2:
Graphical representation of the results of Table 5.1.
It is interesting to repeat this analysis for the full x range. This is done in Fig. 5.3, where R s ( x, Q ) Eq. (5.1) is plotted as a function of x again at low and high scales, now only includingNNPDF3.0, and the default and collider-only versions of NNPDF3.1. It is clear that in thecollider-only PDF set strangeness is largely unconstrained at large x , whereas the global fit isconstrained by neutrino data to have a suppressed value R s ∼ .
5. At lower x we see the tensionbetween this and the constraint from the collider data, which prefer a larger value.In Fig. 5.3 we also compare the strangeness ratio R s ( x, Q ) of NNPDF3.1 with that of CT14and MMHT14. We find that there is good consistency in the entire range of x , while the PDFerrors in NNPDF3.1 are typically smaller than those of the other two sets, especially at largescales. It is also interesting to note how in NNPDF3.1 the PDF uncertainties in the ratio R s blow up at very large x , reflecting the lack of direct information on strangeness in that kinematicregion.We now turn to the strange momentum fraction K s ( Q ) Eq. (5.2); values for the same PDFsets and scales are shown in Table 5.2. Results are quite similar to those found from the analysisof Table 5.1. For the NNPDF3.1 collider-only and especially the HERA + ATLAS W, Z fits,the central value of K s is unphysical, with a huge uncertainty; essentially, all one can say is thatthe strange momentum fraction K s is completely uncertain. This shows rather dramaticallythat the relatively precise values in Table 5.1 only hold in a rather narrow x range. It will beinteresting to see whether more LHC data, possibly leading to a competitive collider-only fit,will confirm strangeness enhancement and allow for an accurate determination of strangeness ina wider range of x . 70 x - - - - ) ( x , Q S R NNPDF3.1 NNPDF3.1 colliderNNPDF3.0
NNLO, Q=1.38 GeV x - - - - ) ( x , Q S R NNPDF3.1 NNPDF3.1 colliderNNPDF3.0
NNLO, Q=100 GeV x - - - - ) ( x , Q S R NNPDF3.1 CT14MMHT14
NNLO, Q=1.38 GeV x - - - - ) ( x , Q S R NNPDF3.1 CT14MMHT14
NNLO, Q=100 GeV
Figure 5.3:
The strangeness ratio R s ( x, Q ) Eq. (5.1) as a function of x for two values of Q , Q = 1 .
38 GeV(left) and Q = m Z (right). Results are shown comparing NNPDF3.1 to NNPDF3.1 and the collider-onlyNNPDF3.1 (top), and to CT14 and MMHT (bottom). The charm content of the proton determined by fitting the charm PDF was quantified within theNNPDF global analysis framework in [23], where results obtained when charm is independentlyparametrized, or perturbatively generated , were compared for the first time. The analysis therewas performed at NLO only, and the dataset was very similar to that of the NNPDF3.0 fit. Wenow re-examine the fitted charm PDF at NNLO in perturbative QCD, and in the context ofthe inclusion of the new datasets, in particular top, and LHCb and ATLAS electroweak bosonproduction, which sizably affects and constrain the charm PDF.Indeed, we have seen in Section 4.1, in particular Fig. 4.2, that the new data added inNNPDF3.1 considerably reduces the charm PDF uncertainty, but also affects its central value,which changes by more than one sigma at large x . Also, in Ref. [23] charm at large x was mostlyconstrained by the EMC data which we discussed in Section 4.9 and which are not included inthe default NNPDF3.1 PDF determination. As we have seen in Section 4.9 these data still havea significant impact on charm, hence a re-assessment of charm determination is in order bothwhen this dataset is included and when it is not. We therefore now compare results obtainedusing the default NNPDF3.1 NNLO set, the modified version of that in which charm is generatedperturbatively as discussed in Section 3.4, and the modified set in which the EMC data are addedto the NNPDF3.1 dataset, as discussed in Section 4.9.In Table 5.3 we show the charm momentum fraction, defined as C ( Q ) ≡ (cid:90) dx (cid:0) xc ( x, Q ) + x ¯ c ( x, Q ) (cid:1) , (5.3)for two values of Q , at the charm threshold Q = m c , and at Q = m Z , computed from using71DF set K s ( Q = 1 .
38 GeV) K s ( Q = M Z )NNPDF3.0 0 . ± .
07 0 . ± . . ± .
07 0 . ± . . ± . . ± . W, Z − . ± . . ± . The strangeness momentum fraction Eq. (5.2) at a low scale and a high scale. We showresults obtained using NNPDF3.0, and NNPDF3.1 baseline, collider-only and HERA+ATLAS
W, Z
PDFsets.
PDF set C ( m c ) C ( m Z )NNPDF3.1 (0 . ± . . ± . . ± . . ± . . ± . . ± . The charm momentum fraction C ( Q ), Eq. (5.3), just above the charm threshold Q = m c GeV and at Q = m Z . Results are shown for NNPDF3.1, and its modified versions in which EMCdata are added to the dataset, or charm is not fitted. these PDF sets. A graphical representation of the results from Table 5.3 is shown in Fig. 5.4.There is a very large difference, by two orders of magnitude, between the uncertainty on themomentum fraction, according to whether charm is independently parametrized, or perturba-tively generated. However, the central values agree with each other within the large uncertaintydetermination when charm is parametrized, with the corresponding central value only slightlylarger than that when charm is perturbative (though hugely different on the scale of the uncer-tainty on the perturbatively generated result). Upon adding the EMC data the uncertainty isreduced by about a factor of three, and the central value somewhat increased, consistently withthe effect of this dataset on the charm PDF discussed in Section 4.9.We can interpret the difference between the total momentum fraction when charm is in-dependently parametrized and determined from the data, and that when charm is perturba-tively generated, as the “intrinsic” (i.e. non-perturbative) charm momentum fraction. In-cluding EMC data when charm is parametrized, at Q = m c we find that it is given by C ( m c ) FC − C ( m c ) PC = (0 . ± . σ level, somewhat improving the estimates of Ref. [23], butstill with a considerable degree of uncertainty. Our estimate for the non-perturbative charmcomponent of the proton is considerably smaller than those allowed in the CT14IC model anal-ysis [136], also shown in Fig. 5.4. However, it is larger than expected in Ref. [137], where anupper bound of 0 .
5% at the four- σ level is claimed. Both these analyses have difficulty fittingthe EMC charm structure function data, due to an overly restrictive functional form for thecharm PDF. At high scale all estimates of C ( Q ) are dominated by the perturbative growth ofthe charm PDF at small x , but the one- σ excess in the fit with EMC data persists, though asan ever diminishing fraction of the whole. This is demonstrated very clearly in Fig. 5.5, wherewe plot the dependence of the charm momentum fraction C ( Q ) on the scale Q .The origin of these values of the charm momentum fraction can be understood by directlycomparing the charm PDF, which is done in Fig. 5.6, again just above threshold Q = m c andat Q = m Z , in the latter case as a ratio to the baseline NNPDF3.1 result. The agreement ofthe charm momentum fraction when it is perturbatively generated or when it is parametrized72 = 1.51 GeV ) [%] c C( Q = m
Momentum Fraction of Charm Quarks
NNPDF3.0NNPDF3.1 NNPDF3.1+EMCCT14IC BHPSCT14IC SEA
Momentum Fraction of Charm Quarks ) [%] Z C( Q = M
Momentum Fraction of Charm Quarks
NNPDF3.0NNPDF3.1 NNPDF3.1+EMCCT14IC BHPSCT14IC SEA
Momentum Fraction of Charm Quarks
Figure 5.4:
Graphical representation of the results for C ( Q ) from Table 5.3 and Q = m c GeV (left)and Q = m Z (right). Model estimates from Ref. [136] are also shown. Q (GeV) M o m en t u m F r a c t i on c a rr i ed b y C ha r m ( % ) Fitted charmFitted charm + EMC dataPerturbative charm
NNPDF3.1 NNLO
Q=m c + ε Figure 5.5:
The charm momentum fraction of Tab. 5.3 plotted as a function of scale Q . and determined from the data is related to the fact that, when parametrized, the best-fit charmhas qualitatively the same shape as charm generated perturbatively at NNLO, as observed inRef. [23] and Section 3.4 above. However, at threshold Q = m c GeV the best-fit charm is largerthan the perturbative component at large x , x ∼ > .
2, albeit with large uncertainties, that aresomewhat reduced when the EMC dataset is added, without a significant change in shape. Uponaddition of the EMC data, in the medium-small 10 − ∼ < x ∼ < − the PDF is pushed at theupper edge of the uncertainty band before addition, with considerably reduced uncertainty. Theunrealistically small uncertainty on the perturbatively generated charm PDF is apparent, andalso the reduction in uncertainty due to the EMC data for all x ∼ > − , already discussed inSection 4.9.The origin of the differences between the charm PDF when perturbatively generated, orparametrized and determined from the data, and a possible decomposition of the latter intoan “intrinsic” (non-perturbative) and a perturbative component can be understood by studyingtheir scale dependence close to threshold, in analogy to a similar analysis presented in Ref. [23].This is done in Fig. 5.7, where the charm PDF is shown (in the n f = 4 scheme) as a function of x both when charm is parametrized and when it is perturbatively generated. On the one hand,the large x ∼ > . x - - - - ) ( x , Q + x c - - - Fitted charmFitted charm + EMCPerturbative charm
NNPDF3.1 NNLO, Q=1.51 GeV x - - - - ) [ r e f] ( x , Q + ) / c ( x , Q + c Z NNPDF3.1 NNLO, Q = M x - - - - ) ( x , Q + c d Fitted charmFitted charm + EMCPerturbative charm
NNPDF3.1 NNLO, Q=1.51 GeV x - - - - ) [ r e f] ) ( x , Q + ) / ( c ( x , Q + c d Fitted charmFitted charm + EMCPerturbative charm Z NNPDF3.1 NNLO, Q = M
Figure 5.6:
Comparison of the charm PDF at the scale and for the PDF sets of Tab. 5.3. Both the PDF(top) and the relative uncertainty (bottom) are shown. charm vanishes identically in this region, so the fitted result in this region may be interpretedas being of non-perturbative origin, i.e. “intrinsic”.On the other hand, for smaller x the charm PDF depends strongly on scale. When pertur-batively generated, it is sizable and positive already at Q ∼ x ∼ > − , while atthreshold it becomes large and negative for all x ∼ < − , possibly because of large unresummedsmall- x contributions. The best-fit parametrized charm PDF, within its larger uncertainty, israther flatter and smaller in modulus essentially for all x ∼ < − , except at the scale-dependentpoint at which perturbative charm changes sign. This difference in shape between fitted charmand perturbative charm for all x ∼ < . (cid:39) .
5% represents a deviation from our best fit value of around two to three sigma.Thus models of intrinsic charm which carry as much as 1% of the proton’s momentum arestrongly disfavored by currently available data.
After analytically discussing the phenomenological implication of individual PDFs and theiruncertainties we now turn to parton luminosities (defined as in Ref. [138]) which drive hadroncollider processes. Parton luminosities from the NNPDF3.0 and NNPDF3.1 NNLO sets for the74 x - - - - = ) f ) ( N x c ( x , Q - - - - - =1.51 GeV cpole NNPDF3.1 NNLO Fitted Charm, m Q = 2.0 GeVQ = 1.7 GeVQ = 1.5 GeVQ = 1.3 GeV=1.51 GeV cpole
NNPDF3.1 NNLO Fitted Charm, m x = ) f ) ( N x c ( x , Q - - =1.51 GeV cpole NNPDF3.1 NNLO Fitted Charm, m Q = 2.0 GeVQ = 1.7 GeVQ = 1.5 GeVQ = 1.3 GeV=1.51 GeV cpole
NNPDF3.1 NNLO Fitted Charm, m x - - - - = ) f ) ( N x c ( x , Q - - - - - =1.51 GeV cpole NNPDF3.1 NNLO Perturbative Charm, m
Q = 2.0 GeVQ = 1.7 GeVQ = 1.5 GeVQ = 1.3 GeV =1.51 GeV cpole
NNPDF3.1 NNLO Perturbative Charm, m x = ) f ) ( N x c ( x , Q - - =1.51 GeV cpole NNPDF3.1 NNLO Perturbative Charm, m
Q = 2.0 GeVQ = 1.7 GeVQ = 1.5 GeVQ = 1.3 GeV =1.51 GeV cpole
NNPDF3.1 NNLO Perturbative Charm, m
Figure 5.7:
The charm PDF in the n f = 4 scheme at small x (left) and large x (right plot) for differentvalues of Q , in the NNPDF3.1 NNLO PDF set (top) and when assuming that charm is perturbativelygenerated (bottom). LHC 13 TeV are compared in Fig. 5.8, and their uncertainties are displayed in Fig. 5.9 as atwo-dimensional contour plot as a function of the invariant mass M y and rapidity y of the finalstate, all normalized to the NNPDF3.1 central value. We show results for the quark-quark,quark-antiquark, gluon-gluon and gluon-quark luminosities, relevant for the measurement offinal states which do not couple to individual flavors (such as Z or Higgs). In the uncertaintyplot we also show for reference the up-antidown luminosity, relevant e.g. for W + production.Two features of this comparison are apparent. First, quark luminosities are generally largerfor all invariant masses, while the gluon luminosity is somewhat enhanced for smaller invariantand somewhat suppressed for larger invariant masses in NNPDF3.1 in comparison to NNPDF3.0.The size of the shift in the quark sector is of order of one sigma or sometimes even larger, whilefor the gluon is generally rather below the one-sigma level. This of course reflects the patternseen in Section 3.3 for PDFs, see in particular Fig. 3.3. Secondly, uncertainties are greatlyreduced in NNPDF3.1 in comparison to NNPDF3.0. This reduction is impressive and apparentin the plots of Fig. 5.9, where it is clear that while uncertainties were typically of order 5%in most of phase space for NNPDF3.0, they are now of the order of 1-2% in a wide centralrapidities range | y | ∼ < ∼ < M x ∼ < ( GeV ) X M10 Q ua r k - Q ua r k Lu m i no s i t y NNPDF3.1NNPDF3.0
LHC 13 TeV, NNLO ( GeV ) X M10 Q ua r k - A n t i qua r k Lu m i no s i t y NNPDF3.1NNPDF3.0
LHC 13 TeV, NNLO ( GeV ) X M10 G l uon - G l uon Lu m i no s i t y NNPDF3.1NNPDF3.0
LHC 13 TeV, NNLO ( GeV ) X M10 Q ua r k - G l uon Lu m i no s i t y NNPDF3.1NNPDF3.0
LHC 13 TeV, NNLO
Figure 5.8:
Comparison of parton luminosities with the NNPDF3.0 and NNPDF3.1 NNLO PDF setsfor the LHC 13 TeV. From left to right and from top to bottom quark-antiquark, quark-quark, gluon-gluon and quark-gluon PDF luminosities are shown. Results are shown normalized to the central valueof NNPDF3.1.
Agreement becomes marginal at large masses, M X ∼ > x PDFs. For the gluon-gluon and gluon-quark channels we find reasonable agreementfor masses up to M X (cid:39)
600 GeV, relevant for precision physics at the LHC, but rather worseagreement for larger masses, relevant for BSM searches, in particular between NNPDF3.1 andMMHT14. Of course it should be kept in mind that NNPDF3.1 has a wider dataset and a largernumber of independently parametrized PDFs than MMHT14 and CT14, hence the situation maychange in the future once all global PDF sets are updated.Next, in Fig. 5.11 we compare to ABMP16 PDFs. In this case, we show results correspondingboth to the default ABMP16 set, which has α s ( m Z ) = 0 . α s ( m Z ) = 0 .
118 adopted so far in all comparison. While there are sizable differences betweenNNPDF3.1 and ABMP16 when the default ABMP16 value α s ( m Z ) = 0 . α s ( m Z ) = 0 .
118 is adopted alsofor ABMP16. However, ABMP16 luminosities have very small uncertainties at low and high M X , presumably a consequence of an over-constrained parametrization, and of using a Hessianapproach but with no tolerance, as discussed in Section 3.3.76 M X ( G e V ) Relative uncertainty for qq-luminosityNNPDF3.0 NNLO - s = 13000.0 GeV R e l a t i v e un c e r t a i n t y ( % ) M X ( G e V ) Relative uncertainty for qq-luminosityNNPDF 3.1 NNLO - s = 13000.0 GeV R e l a t i v e un c e r t a i n t y ( % ) M X ( G e V ) Relative uncertainty for qq-luminosityNNPDF30_nnlo_as_0118 - s = 13000.0 GeV R e l a t i v e un c e r t a i n t y ( % ) M X ( G e V ) Relative uncertainty for qq-luminosityNNPDF 3.1 NNLO - s = 13000.0 GeV R e l a t i v e un c e r t a i n t y ( % ) M X ( G e V ) Relative uncertainty for gg-luminosityNNPDF3.0 NNLO - s = 13000.0 GeV R e l a t i v e un c e r t a i n t y ( % ) M X ( G e V ) Relative uncertainty for gg-luminosityNNPDF 3.1 NNLO - s = 13000.0 GeV R e l a t i v e un c e r t a i n t y ( % ) M X ( G e V ) Relative uncertainty for gq-luminosityNNPDF3.0 NNLO - s = 13000.0 GeV R e l a t i v e un c e r t a i n t y ( % ) M X ( G e V ) Relative uncertainty for gq-luminosityNNPDF 3.1 NNLO - s = 13000.0 GeV R e l a t i v e un c e r t a i n t y ( % ) M X ( G e V ) Relative uncertainty for ud-luminosityNNPDF3.0 NNLO - s = 13000.0 GeV R e l a t i v e un c e r t a i n t y ( % ) M X ( G e V ) Relative uncertainty for ud-luminosityNNPDF31_nnlo_as_0118 - s = 13000.0 GeV R e l a t i v e un c e r t a i n t y ( % ) Figure 5.9:
The relative uncertainty on the luminosities of Fig. 5.8, plotted as a function of the invariantmass M X and the rapidity y of the final state; the left plots show results for NNPDF3.0 and the rightplots for NNPDF3.1 (upper four rows). The bottom row shows results for the up-antidown luminosity. ( GeV ) X M10 Q ua r k - Q ua r k Lu m i no s i t y NNPDF3.1CT14MMHT14
LHC 13 TeV, NNLO ( GeV ) X M10 Q ua r k - A n t i qua r k Lu m i no s i t y NNPDF3.1CT14MMHT14
LHC 13 TeV, NNLO ( GeV ) X M10 G l uon - G l uon Lu m i no s i t y NNPDF3.1CT14MMHT14
LHC 13 TeV, NNLO ( GeV ) X M10 Q ua r k - G l uon Lu m i no s i t y NNPDF3.1CT14MMHT14
LHC 13 TeV, NNLO
Figure 5.10:
Same as Fig. 5.8, now comparing NNPDF3.1 NNLO to CT14 and MMHT14.
Finally, in order to further emphasize the phenomenological impact of the new NNPDF3.1methodology, we compare NNPDF3.1 PDFs to the modified version in which the charm PDFis perturbatively generated, already discussed in Sects. 3.4, 5.3. Results are shown in Fig. 5.12.On the one hand, we confirm that despite having one more parametrized PDF, uncertainties arenot increased. On the other hand, the effect on central values is moderate but non-negligible.For gluon-gluon and quark-gluon luminosities, differences are always below the one-sigma level,and typically rather less. For the quark-quark channel, results do not depend on the charmtreatment for M X ∼ >
200 GeV, but for smaller invariant masses perturbatively generated charmleads to a larger PDF luminosity than the best-fit parametrized charm. For the quark-antiquarkluminosity, we find a similar pattern at small M X , but also some differences at medium andlarge M X . 78 ( GeV ) X M10 Q ua r k - Q ua r k Lu m i no s i t y NNPDF3.1 =0.118 S a ABMP16 =0.1147 S a ABMP16
LHC 13 TeV, NNLO ( GeV ) X M10 Q ua r k - A n t i qua r k Lu m i no s i t y NNPDF3.1 =0.118 S a ABMP16 =0.1147 S a ABMP16
LHC 13 TeV, NNLO ( GeV ) X M10 G l uon - G l uon Lu m i no s i t y NNPDF3.1 =0.118 S a ABMP16 =0.1147 S a ABMP16
LHC 13 TeV, NNLO ( GeV ) X M10 Q ua r k - G l uon Lu m i no s i t y NNPDF3.1 =0.118 S a ABMP16 =0.1147 S a ABMP16
LHC 13 TeV, NNLO
Figure 5.11:
Same as Fig. 5.10, now comparing to ABMP16 PDFs, both with their default α s ( m Z ) =0 . α s ( m Z ) = 0 . ( GeV ) X M10 Q ua r k - Q ua r k Lu m i no s i t y fitted charmperturbative charm LHC 13 TeV, NNLO ( GeV ) X M10 Q ua r k - A n t i qua r k Lu m i no s i t y fitted charmperturbative charm LHC 13 TeV, NNLO ( GeV ) X M10 G l uon - G l uon Lu m i no s i t y fitted charmperturbative charm LHC 13 TeV, NNLO ( GeV ) X M10 Q ua r k - G l uon Lu m i no s i t y fitted charmperturbative charm LHC 13 TeV, NNLO
Figure 5.12:
Same as Fig. 5.8, now comparing NNPDF3.1 to its modified version with perturbativecharm. .
26 1 .
27 1 .
28 1 .
29 1 .
30 1 .
31 1 .
32 1 . σ W + /σ W − ATLAS 13 TeV
Heavy: NNLO QCD + NLO EWLight: NNLO QCD
Ratio of W + to W − boson NNPDF3.1NNPDF3.0CT14 MMHT14ABMP16data ± total uncertainty . . . . . . σ W /σ Z Heavy: NNLO QCD + NLO EWLight: NNLO QCD
ATLAS 13 TeV
Ratio of W ± to Z boson NNPDF3.1NNPDF3.0CT14 MMHT14ABMP16data ± total uncertainty Figure 5.13:
Comparison of the ATLAS measurements of the W + /W − ratio (left) and the W/Z ratio(right) at √ s = 13 TeV with theoretical predictions computed with different NNLO PDF sets. Predictionsare shown with (heavy) and without (light) NLO EW corrections computed with FEWZ and
HORACE , asdescribed in the text. W and Z production at the LHC 13 TeV We compare theoretical predictions based on the NNPDF3.1 set to W and Z production data at √ s = 13 TeV from ATLAS [139]. Similar measurements by CMS [140] are not included in thiscomparison as they are still preliminary. We compute fiducial cross-sections using FEWZ [114]at NNLO QCD accuracy, using NNPDF3.1, NNPDF3.0, CT14, MMHT14 and ABMP16 PDFs,together with the corresponding PDF uncertainty band. All calculations (including ABMP16)are performed with α s = 0 . FEWZ for Z production, and with HORACE3.2 [141] for W production. The fiducial phase space for the W ± cross-section measurement in ATLAS is by p lT ≥
25 GeV and | η l | ≤ . p νT ≥
25 GeV and a W transversemass of m T ≥
50 GeV. For Z production, p lT ≥
25 GeV and | η l | ≤ . ≤ m ll ≤
116 GeV for the dilepton invariant mass.In Fig. 5.13 we compare the ATLAS [139] 13 TeV measurements of the the W + /W − and W/Z ratios in the fiducial region at √ s = 13 TeV to theoretical predictions, both with andwithout electroweak corrections. We see that for both the W + /W − ratio and the W/Z ratio,all the PDF sets are in reasonably good agreement with the data. The uncertainty in thetheoretical prediction shown in the plot is the PDF uncertainty only: parametric uncertainties(in the values of α s and m c ) and missing higher order QCD uncertainties are not included.Interestingly, electroweak corrections shift the theory predictions by around 0.5%, and for allPDF sets they improve the agreement with the ATLAS measurements. NNPDF3.1 results havesmaller PDF uncertainties than NNPDF3.0, and are in better agreement with the ATLAS data.The corresponding absolute W + , W − and Z cross-sections are shown in Fig. 5.14, normalizedin each case to the experimental central value. Again, predictions are generally in agreementwith the data, with the possible exception of ABMP16 for Z production. In comparison tocross-section ratios, the effect of electroweak corrections on absolute cross-sections, around 1%for W + and W − and around 0.5% for Z production, is rather less significant on the scale ofthe uncertainties involved, and it does not necessarily lead to improved agreement. Compar-ing NNPDF3.1 to NNPDF3.0 we see again considerably reduced uncertainties and improvedagreement of the prediction with data. This improved agreement is particularly marked for Z production, where 3.0 was about 5% below the data, while now 3.1 agrees within uncertainties.80 . . . . . . σ p r e d . / σ m e a s . ATLAS 13 TeV
Heavy: NNLO QCD + NLO EWLight: NNLO QCD W − W + Z NNPDF3.1NNPDF3.0CT14 MMHT14ABMP16data ± total uncertainty Figure 5.14:
Same as Fig. 5.13, now for the absolute W + , W − and Z cross-sections. All predictions arenormalized to the experimental central value. We finally study the PDF dependence of predictions for inclusive Higgs production at LHC13 TeV, and for Higgs pair production, which could also be within reach of the LHC in thenear future [142, 143]. We study single Higgs in gluon fusion, associated production with gaugebosons and top pairs and vector boson fusion, and double Higgs production in gluon fusion.In each case, we show predictions normalized to the NNPDF3.1 result, and only show PDFuncertainties. All calculations (including ABMP16) are performed with α s = 0 . LO using ggHiggs [144–146]. Renormalization and factorization scales are set to µ F = µ R = m h / t ¯ t pair we use MadGraph5 aMC@NLO [147], with default factorization and renormalization scales µ R = µ F = H T /
2, where H T is the sum of the transverse masses. For associate production withan electroweak gauge boson we use vh@nnlo code [148] at NNLO with default scale settings.For vector boson fusion we perform the calculation at N LO using proVBFH [149, 150] with thedefault scale settings. Finally, for double Higgs production at the FCC 100 TeV the calculationis performed using
MadGraph aMC@NLO .Results are shown in Figs. 5.15-5.16. For gluon fusion and t ¯ th , which are both driven by thegluon PDF, the former for x ∼ − , and the latter for large x , results from the various PDFsets agree within uncertainties; NNPDF3.0 and NNPDF3.1 are also in good agreement, withthe new prediction exhibiting reduced uncertainties. The spread of results is somewhat largerfor associate production with gauge bosons. The NNPDF3.1 prediction is about 3% higher thanthe NNPDF3.0 one, with uncertainties reduced by a factor 2, so the two cross-sections barelyagree within uncertainties. Also, of the three PDF sets entering the PDF4LHC15 combination,NNPDF3.0 gave the smallest cross-section, but NNPDF3.1 now gives the highest one: V H production is driven by the quark-antiquark luminosity, and this enhancement for M X (cid:39) .
90 0 .
92 0 .
94 0 .
96 0 .
98 1 .
00 1 .
02 1 . m H = 125 GeV Higgs production: gluon fusion
NNPDF3.1NNPDF3.0CT14MMHT14ABMP16 .
80 0 .
85 0 .
90 0 .
95 1 .
00 1 .
05 1 . m H = 125 GeV Higgs production: associate production with t ¯ t NNPDF3.1NNPDF3.0CT14MMHT14ABMP16 .
90 0 .
92 0 .
94 0 .
96 0 .
98 1 .
00 1 .
02 1 . m H = 125 GeVheavy: W + productionlight: W − production Higgs production:
W H associate production
NNPDF3.1NNPDF3.0CT14 MMHT14ABMP16 .
90 0 .
92 0 .
94 0 .
96 0 .
98 1 .
00 1 .
02 1 . m H = 125 GeV Higgs production: ZH associate production NNPDF3.1NNPDF3.0CT14 MMHT14ABMP16
Figure 5.15:
PDF dependence of the Higgs production cross-sections at the LHC 13 TeV for gluonfusion, t ¯ t associated production, and V H associated production. All results are shown as ratios to thecentral NNPDF3.1 result. Only PDF uncertainties are shown.
NNPDF3.1 is the new main PDF release from the NNPDF family. It represents a significant im-provement over NNPDF3.0, by including constraints from many new observables, some of whichare included for the first time in a global PDF determination, thanks to the recent availabilityof the corresponding NNLO QCD corrections. Notable examples are t ¯ t differential distributionsand the Z boson p T spectrum. From the theory point of view, the main improvement is to placethe charm PDF on an equal footing as the light quark PDFs. Independently parametrizing thecharm PDF resolves a tension which would otherwise be present between ATLAS gauge bosonproduction and HERA inclusive structure function data, leads to improved agreement with theLHC data, and turns the strong dependence of perturbatively generated charm on the value ofthe pole charm mass into a PDF uncertainty, as most of the mass dependence is reabsorbed intothe initial PDF shape.The NNPDF3.1 set is also the first set for which PDFs are delivered in a variety of formats:first of all, they are released both in Hessian and Monte Carlo form, and furthermore, the defaultsets are optimized and compressed so that a smaller number of Monte Carlo replicas or Hessianerror sets reproduces the statistical features of much larger underlying replica sets. We nowdiscuss how both Hessian and Monte Carlo reduced sets have been produced out of a large setof Monte Carlo replicas; we then summarize all PDF sets that have been made public throughthe LHAPDF interface; and finally we present a brief outlook on future developments.82 .
90 0 .
92 0 .
94 0 .
96 0 .
98 1 .
00 1 .
02 1 . m H = 125 GeV Higgs production: Vector Boson Fusion
NNPDF3.1NNPDF3.0CT14MMHT14ABMP16 .
90 0 .
92 0 .
94 0 .
96 0 .
98 1 .
00 1 .
02 1 . m H = 125 GeV Double Higgs production
NNPDF3.1NNPDF3.0CT14MMHT14ABM16
Figure 5.16:
Same as Fig. 5.15 for single Higgs production in vector boson fusion (left) and double Higgsproduction in gluon fusion (right).
Default NNPDF3.1 NLO and NNLO PDFs for the central α s ( m Z ) = 0 .
118 value, as well asthe modified version with perturbative charm discussed in Sect. 3.4, have been produced as N rep = 1000 replica sets. These large replica samples have been subsequently processed usingtwo reduction strategies: the Compressed Monte Carlo (CMC) algorithm [25], to obtain a MonteCarlo representation based on a smaller number of replicas, and the MC2H algorithm [24], toachieve an optimal Hessian representation of the underlying PDF probability distribution with afixed number of error sets. Specifically, we have thus constructed CMC-PDF sets with N rep = 100replicas and MC2H sets with N eig = 100 (symmetric) eigenvectors.In Fig. 6.1 we show the comparison between the PDFs from the input set of N rep = 1000replicas of NNPDF3.1 NNLO with the corresponding reduced sets of the CMC-PDFs with N rep = 100 replicas and the MC2H hessian PDFs with N eig = 100 eigenvalues. The agreementbetween the input N rep = 1000 replica MC PDFs and the two reduced sets is very good in allcases. By construction, the agreement in central values and PDF variances is slightly betterfor the MC2H sets, since the CMC-PDF sets aim to reproduce also higher moments in theprobability distribution and thus possibly non-gaussian features, while Hessian sets are Gaussianby construction. Following the analysis of [24, 25] we have verified that also the correlationsbetween PDFs are reproduced to a high degree of accuracy.In order to validate the efficiency of the CMC-PDF algorithm reduction from the starting N rep = 1000 replicas down to the compressed N rep = 100 replicas, in Fig. 6.2 we show, followingthe procedure described in [25], the summary of statistical estimators that compare specific prop-erties of the probability distributions defined by the input N rep = 1000 replicas of NNPDF3.1NNLO and the corresponding compressed sets as a function of (cid:101) N rep , the number of replicas inthe reduced set starting from (cid:101) N rep = 100. We compare the results of the compression algorithmwith those of random selection of (cid:101) N rep replicas out of the original 1000 ones: the error functionERF corresponding to central values, standard deviations, kurtosis and skewness, correlationsand the Kolmogorov distance are all shown. These results indicate that a CMC-PDF 100 replicaset reproduces roughly the information contained in a random (cid:101) N rep = 400 PDF set.83 x - - - - ) [ r e f] ) / u ( x , Q u ( x , Q =1000 rep Input MC, N =100 rep
Compressed MC, N=100 eig
Hessian, N GeV =10 NNPDF3.1 NNLO, Q x - - - - ) [ r e f] ) / d ( x , Q d ( x , Q =1000 rep Input MC, N =100 rep
Compressed MC, N=100 eig
Hessian, N GeV =10 NNPDF3.1 NNLO, Q x - - - - ) [ r e f] ( x , Q u ) / ( x , Q u =1000 rep Input MC, N =100 rep
Compressed MC, N=100 eig
Hessian, N GeV =10 NNPDF3.1 NNLO, Q x - - - - ) [ r e f] ( x , Q d ) / ( x , Q d =1000 rep Input MC, N =100 rep
Compressed MC, N=100 eig
Hessian, N GeV =10 NNPDF3.1 NNLO, Q x - - - - ) [ r e f] ( x , Q + ) / s ( x , Q + s =1000 rep Input MC, N =100 rep
Compressed MC, N=100 eig
Hessian, N GeV =10 NNPDF3.1 NNLO, Q x - - - - ) [ r e f] ( x , Q + ) / c ( x , Q + c =1000 rep Input MC, N =100 rep
Compressed MC, N=100 eig
Hessian, N GeV =10 NNPDF3.1 NNLO, Q x - - - - ) [ r e f] ( x , Q S ) / ( x , Q S =1000 rep Input MC, N =100 rep
Compressed MC, N=100 eig
Hessian, N GeV =10 NNPDF3.1 NNLO, Q x - - - - ) [ r e f] ) / g ( x , Q g ( x , Q =1000 rep Input MC, N =100 rep
Compressed MC, N=100 eig
Hessian, N GeV =10 NNPDF3.1 NNLO, Q
Figure 6.1:
Comparison between the PDFs from the input set of N rep = 1000 replicas of NNPDF3.1NNLO, the reduced Monte Carlo CMC-PDFs with N rep = 100 replicas, and the MC2H hessian PDFswith N eig = 100 symmetric eigenvalues. eplicas ERF Central Value - NNPDF3.1 NNLO
CompressedRandom Mean (1k)Random Median (1k)Random 50% c.l. (1k)Random 68% c.l. (1k)Random 90% c.l. (1k)
Replicas -
10 110
ERF Standard deviation - NNPDF3.1 NNLO
CompressedRandom Mean (1k)Random Median (1k)Random 50% c.l. (1k)Random 68% c.l. (1k)Random 90% c.l. (1k)
Replicas ERF Skewness - NNPDF3.1 NNLO
CompressedRandom Mean (1k)Random Median (1k)Random 50% c.l. (1k)Random 68% c.l. (1k)Random 90% c.l. (1k)
Replicas ERF Kurtosis - NNPDF3.1 NNLO
CompressedRandom Mean (1k)Random Median (1k)Random 50% c.l. (1k)Random 68% c.l. (1k)Random 90% c.l. (1k)
Replicas ERF Kolmogorov - NNPDF3.1 NNLO
CompressedRandom Mean (1k)Random Median (1k)Random 50% c.l. (1k)Random 68% c.l. (1k)Random 90% c.l. (1k)
Replicas - - - - - - - ERF Correlation - NNPDF3.1 NNLO
CompressedRandom Mean (1k)Random Median (1k)Random 50% c.l. (1k)Random 68% c.l. (1k)Random 90% c.l. (1k)
Figure 6.2:
Comparison of estimators of the probability distributions computed using the NNPDF3.1NNLO input N rep = 1000 replica set, and compressed sets of (cid:101) N rep replicas, plotted as a function of (cid:101) N rep . The error function (ERF) corresponding to central values, standard deviations, kurtosis, skewness,correlations and Kolmogorov distance are shown. .2 Delivery We now provide a full list of the NNPDF3.1 PDF sets that are being made publicly availablevia the
LHAPDF6 interface [151], http://lhapdf.hepforge.org/ .As repeatedly mentioned in the paper, a very wide set of results concerning these PDF sets isavailable from the repository http://nnpdf.hepforge.org/html/nnpdf31/catalog .All sets are made available as N rep = 100 Monte Carlo sets. For the baseline sets, these areconstructed out of larger N rep = 1000 replica sets, which are also being made available. Thebaseline sets are also provided as Hessian sets with 100 error sets.The full list is the following: • Baseline NLO and NNLO NNPDF3.1 sets
Baseline NLO and NNLO NNPDF3.1 sets are based on the global dataset, with α s ( m Z ) =0 .
118 and a variable-flavor number with up to five active flavors. These sets contain N rep = 1000 PDF replicas. NNPDF31 nlo as 0118 1000NNPDF31 nnlo as 0118 1000.
A modified version in which charm is perturbatively generated (ad in previous NNPDFsets) is also being made available, also with N rep = 1000 PDF replicas: NNPDF31 nlo pch as 0118 1000NNPDF31 nnlo pch as 0118 1000.
Out of these , optimized Monte Carlo N rep = 100 NNPDF31 nlo as 0118NNPDF31 nnlo as 0118NNPDF31 nlo pch as 0118NNPDF31 nnlo pch as 0118, and Hessian sets with N eig = 100 eigenvectors NNPDF31 nlo as 0118 hessianNNPDF31 nnlo as 0118 hessianNNPDF31 nlo pch as 0118 hessianNNPDF31 nnlo pch as 0118 hessian have been constructed as discussed in Sect. 6.1 above.Out of these, smaller sets of eigenvectors optimized for the computation of specific observ-ables may be constructed using the
SM-PDF tool [26]. Specifically, sets optimized for a widelist of predefined observables can be generated and downloaded using the web interface [27]at https://smpdf.mi.infn.it .86
Flavor number variation
We have produced sets, both at NLO and NNLO in which the maximum number of flavorsdiffers from the default 5, and it is either extended up to six, or frozen at four:
NNPDF31 nlo as 0118 nf 4NNPDF31 nlo as 0118 nf 6NNPDF31 nnlo as 0118 nf 4NNPDF31 nnlo as 0118 nf 6.
The variant with perturbatively generated charm is also made available, in this case alsoin a n f = 3 fixed-flavor number scheme: NNPDF31 nlo pch as 0118 nf 3NNPDF31 nlo pch as 0118 nf 4NNPDF31 nlo pch as 0118 nf 6NNPDF31 nnlo pch as 0118 nf 3NNPDF31 nnlo pch as 0118 nf 4NNPDF31 nnlo pch as 0118 nf 6. • α s variation. We have produced NNLO sets with the following values of α s ( m Z ) 0.108, 1.110, 0.112,0.114, 0.116, 0.117, 0.118, 0.119, 0.120, 0.122, 0.124: NNPDF31 nnlo as 0108NNPDF31 nnlo as 0110NNPDF31 nnlo as 0112NNPDF31 nnlo as 0114NNPDF31 nnlo as 0116NNPDF31 nnlo as 0117NNPDF31 nnlo as 0118NNPDF31 nnlo as 0119NNPDF31 nnlo as 0120NNPDF31 nnlo as 0122NNPDF31 nnlo as 0124.
For the values α s ( m Z ) = 0 .
116 and α s ( m Z ) = 0 .
120 we have also produced NLO sets,
NNPDF31 nlo as 0116NNPDF31 nlo as 0120, and the variant with perturbative charm both at NLO and NNLO
NNPDF31 nlo pch as 0116NNPDF31 nlo pch as 0120NNPDF31 nnlo pch as 0116NNPDF31 nnlo pch as 0120.
In order to facilitate the computation of the combined PDF+ α s uncertainties we have alsoprovided bundled PDF+ α s variation sets for α s ( m Z ) = 0 . ± . NPDF31 nlo pdfasNNPDF31 nnlo pdfasNNPDF31 nlo pch pdfasNNPDF31 nnlo pch pdfasNNPDF31 nlo hessian pdfasNNPDF31 nnlo hessian pdfasNNPDF31 nlo pch hessian pdfasNNPDF31 nnlo pch hessian pdfas.
They are constructed as follows:1. The central value (PDF member 0) is the central value of the corresponding α s ( m Z ) =0 .
118 set.2. The PDF members 1 to 100 correspond to the N rep = 100 ( N eig = 100) Monte Carloreplicas (Hessian eigenvectors) from the α s ( m Z ) = 0 .
118 set.3. The PDF members 101 and 102 are the central values of the sets with α s ( m Z ) = 0 . α s ( m Z ) = 0 .
120 respectively.Note that, therefore, in the Hessian case member 0 is the central set, and all remainingbundled members 1 −
102 are error sets, while in the Monte Carlo case, members 1 − α s uncertainties is discussede.g. in Ref. [12]. • Charm mass variation
We provide sets with different values of the charm mass m pole c . They are available only atNNLO with m pole c = 1 .
38 GeV and m pole c = 1 .
64 GeV.
NNPDF31 nnlo as 0118 mc 138NNPDF31 nnlo as 0118 mc 164 .For comparison, the corresponding modified version with perturbative charm are also madeavailable:
NNPDF31 nnlo pch as 0118 mc 138NNPDF31 nnlo pch as 0118 mc 164 . • Forced positivity sets
We provide sets in which PDFs are non-negative:
NNPDF31 nlo as 0118 mcNNPDF31 nlo pch as 0118 mcNNPDF31 nnlo as 0118 mcNNPDF31 nnlo pch as 0118 mc
These have been constructed simply setting to zero PDFs whenever they become negative.They are thus an approximation, provided for convenience for use in conjunction withcodes which fail when PDFs are negative. • LO sets
Leading-order PDF sets are made available α s = 0 .
118 and α s = 0 . NPDF31 lo as 0118NNPDF31 lo as 0130
The corresponding variant with perturbative charm is also provided:
NNPDF31 lo pch as 0118NNPDF31 lo pch as 0130 . • Reduced datasets
PDFs determined from subsets of the full NNPDF3.1, discussed in Sect. 4.2-4.12 are alsomade available, specifically
NNPDF31 nnlo as 0118 noZpt no Z p T data, see Sect. 4.2; NNPDF31 nnlo as 0118 notop no t ¯ t production data, see Sect. 4.3; NNPDF31 nnlo as 0118 nojets no jet data, see Sect. 4.4;
NNPDF31 nnlo as 0118 wEMC
EMC charm data added, see Sect. 4.9;
NNPDF31 nnlo as 0118 noLHC no LHC data, see Sect. 4.10;
NNPDF31 nnlo as 0118 proton proton-target data only, see Sect. 4.11;
NNPDF31 nnlo as 0118 collider collider data only, see Sect. 4.12.In addition to these, any other PDF sets discussed in this paper is also available upon request.
The NNPDF3.1 PDF determination presented here is an update of the previous NNPDF3.0,yet it contains substantial innovations both in terms of methodology and dataset and it leadsto substantially more precise and accurate PDF sets. Thanks to this, several spin-offs can bepursued, either with the goal of updating existing results based on previous NNPDF releases,or in some cases because the greater accuracy enables projects which previously were eitherimpossible or uninteresting.These spin-off projects include the following: • A precision determination of the strong coupling constant α s ( m Z ). We expect a significantincrease in both precision and accuracy compared to previous NNPDF determinations [152,153]. Specifically, thanks to the inclusion of many collider observables at NNLO with directdependence on both the gluon and on α s ( m Z ) it might turn out to be advantageous to dropaltogether data taken on nuclear targets, and at low scales, i.e. base the α s determinationon the collider-only dataset of Sect. 4.12, or possibly an even more conservative dataset. • A determination of the charm mass m c . This would be for the first time based on a PDFdetermination in which the charm PDF is independently parametrized, thereby avoidingbias related to the identification of m c as a parameter which determines the size of thecharm PDF. • A NNPDF3.1QED PDF set including the photon PDF γ ( x, Q ), thereby updating theprevious NNPDF2.3QED [154] and NNPDF3.0QED [155] sets. This update should includerecent theoretical progress: specifically the direct “LuxQED” constraints which determinethe photon PDF from nucleon structure functions [156]; and NLO QED corrections toPDF evolution and DIS coefficient functions, now included in APFEL [99].89
PDF sets including small- x resummation, based on the formalism developed in [157–159],which has been implemented in the HELL code [160] and interfaced to
APFEL . These wouldprovide an answer the long-standing issue of whether or not small- x inclusive HERAdata are adequately described by fixed-order perturbative evolution [161, 162], and mayultimately give better control of theoretical uncertainties at small x .On a longer timescale, further substantial improvements in dataset and methodology areexpected. On the one hand, so far we have essentially restricted ourselves to LHC 8 TeV data.A future release will include a significant number of 13 TeV measurements, of which several,from ATLAS, CMS and LHCb are already available. Specifically, the inclusion of more processesis envisaged, which are not currently part of the dataset, which have a large potential impacton PDFs, and for which higher order corrections have become available . These include promptphoton production [163], for which NNLO corrections are now available [164] and whose impacton PDF is well-known [165]; single-top production (also known at NNLO [166]); and possiblyforward D meson production or more in general processes with final-state D mesons, such as W + D production, recently measured by ATLAS [167], a process whose impact on PDFs hasbeen repeatedly emphasized [168–170].Such a further increase in dataset is likely to require substantial methodological improve-ments in PDF determination. Also, it is likely to result in a further reduction of PDF un-certainties, thereby requiring better control of theoretical uncertainties. We specifically expectsignificant progress in two different directions. On the one hand, electroweak corrections, whichare now not included, and whose impact is kept under control through kinematic cuts, will haveto be included in a more systematic way. On the other hand, theoretical uncertainties due tomissing higher-order corrections will also have to be estimated. Indeed, the preliminary esti-mates presented in Sect. 3.5 suggest that missing higher-order uncertainties on PDFs, currentlynot included in the PDF uncertainty, are likely to soon become non-negligible, and possiblydominant in some kinematic regions. All of these improvements will be part of a future majorPDF release. Acknowledgments
We thank M. Bonvini, A. de Roeck, L. Harland-Lang, J. Huston, J. Gao, P. Nadolsky, andR. Thorne for many illuminating discussions. We thank U. Blumenschein, M. Boonekamp,T. Carli, S. Camarda A. Cooper-Sarkar, J. Kretzschmar, K. Lohwasser, K. Rabbertz, V. Radescuand M. Schott for help with the ATLAS measurements. We are grateful to A. Giammanco,R. Gonzalez-Suarez, I. Josa, K. Lipka, M. Owen, L. Perrozzi, R. Placakyte and P. Silva forassistance with the CMS data. We thank S. Farry, P. Ilten and R. McNulty for help with theLHCb data. We thank F. Petriello and R. Boughezal for providing us the NNLO calculationsof the
Z p T distributions. We thank M. Czakon and A. Mitov for assistance and discussions forthe NNLO top quark pair differential distributions. We thank J. Currie, N. Glover and J. Piresfor providing us with the NNLO K -factors for inclusive jet production at 7 TeV and for manydiscussions about jet data in PDF determination. We thank F. Dreyer and A. Karlberg forsending us a private version of the proVBFH code. We thank R. Harlander for assistance with vh@nnlo . We thank L. Harland-Lang for benchmarking the NNLO calculation of the CMS 8TeV Drell-Yan measurements. 90. B., N. H., J. R., L. R. and E. S. are supported by an European Research Council StartingGrant “PDF4BSM”. R. D. B. and L. D. D. are supported by the UK STFC grants ST/L000458/1and ST/P000630/1. L. D. D. is supported by the Royal Society, Wolfson Research Merit Award,grant WM140078. S. F. is supported by the European Research Council under the GrantAgreement 740006NNNPDFERC-2016-ADG/ERC-2016-ADG. E. R. N. is supported by the UKSTFC grant ST/M003787/1. S. C. is supported by the HICCUP ERC Consolidator grant(614577). M. U. is supported by a Royal Society Dorothy Hodgkin Research Fellowship andpartially supported by the STFC grant ST/L000385/1. S. F and Z. K. are supported by theExecutive Research Agency (REA) of the European Commission under the Grant AgreementPITN-GA-2012-316704 (HiggsTools). A. G. is supported by the European Union’s Horizon2020 research and innovation programme under the Marie Sk(cid:32)lodowska-Curie grant agreementNo 659128 - NEXTGENPDF. A Code development and benchmarking
In all previous NNPDF releases, including NNPDF3.0, PDF evolution and deep-inelastic scatter-ing were computed using the internal
FKgenerator code. As discussed in Sect. 2, in NNPDF3.1PDF evolution is now performed using the public
APFEL code. Deep-inelastic scattering andfixed-target Drell Yan are also computed using
APFEL . As far as perturbative evolution isconcerned,
APFEL has been extensively tested against other publicly available codes, such as
HOPPET [109] and
QCDNUM [171] for the PDF evolution and
OpenQCDrad [172] for the calculationof heavy-quark structure functions in the massive scheme.We have performed an extensive benchmarking of
APFEL and
FKgenerator , also involvingdeep-inelastic structure function (as already mentioned in Ref. [23]) and Drell-Yan cross-sections.In the process of this benchmarking, two bugs were found in the
FKgenerator implementation ofthe DIS structure functions (one related to target-mass corrections, the other in the expressionsof the O ( α s ) charge-current massive coefficient functions): we checked explicitly that none ofthem produced an effect on NNPDF3.0 PDFs that could be distinguished from a statisticalfluctuation.Representative results of the benchmarking of deep-inelastic structure functions are shown inFig. A.1, where we show the relative difference between FKgenerator and
APFEL implementation,using NNPDF3.0 as input PDF set, for the CHORUS charged current neutrino-nucleus reducedcross-sections, the NMC proton reduced cross-sections and neutral and charged current cross-sections from the H1 experiments from the HERA-II dataset. In each case, we show theoreticalpredictions calculations at LO and in the FONLL-A, -B and -C general-mass schemes. The twocodes are in good agreement, with differences at most being at the 1% level, typically muchsmaller. This statement holds for all perturbative orders.The benchmarking of Drell-Yan cross-sections is illustrated in Fig. A.2, where we show therelative difference between the
FKgenerator and
APFEL calculations at LO, NLO and NNLO forthe E605 pd and E866 pp cross-sections, again using NNPDF3.0 as input. Again, differences areat most at the 2% level and typically much smaller. We have traced these residual differenceswere to the fact that the coverage of the large- x region was sub-optimal in the FKgenerator calculation, and is now improved in
APFEL thanks to of a better choice of input x grid.91igure A.1: Relative difference in the DIS structure functions computed with
FKgenerator and
APFEL ,using NNPDF3.0 as input PDF set, for the CHORUS charged current neutrino-nucleus reduced cross-sections, the NMC proton reduced cross-sections and neutral and charged current cross-sections fromthe H1 experiments from the HERA-II dataset. Datasets are as in Tab. 1 of Ref. [5]. For each dataset,we compare the theoretical calculations at LO and in the FONLL-A, B and C [100] heavy-quark massschemes.
Figure A.2:
Same as Fig. A.1 for fixed-target Drell-Yan cross-sections: results are shown at LO, NLOand NNLO for the E605 pA and E866 pp cross-sections datasets, as given in Tab. 2 of Ref. [5]. eferences [1] S. Forte, Parton distributions at the dawn of the LHC , Acta Phys.Polon.
B41 (2010)2859, [ arXiv:1011.5247 ].[2] S. Forte and G. Watt,
Progress in the Determination of the Partonic Structure of theProton , Ann.Rev.Nucl.Part.Sci. (2013) 291, [ arXiv:1301.6754 ].[3] R. D. Ball, Global Parton Distributions for the LHC Run II , Nuovo Cim.
C38 (2016),no. 4 127, [ arXiv:1507.07891 ].[4] J. Rojo et al.,
The PDF4LHC report on PDFs and LHC data: Results from Run I andpreparation for Run II , J. Phys.
G42 (2015) 103103, [ arXiv:1507.00556 ].[5]
NNPDF
Collaboration, R. D. Ball et al.,
Parton distributions for the LHC Run II , JHEP (2015) 040, [ arXiv:1410.8849 ].[6] S. Dulat, T.-J. Hou, J. Gao, M. Guzzi, J. Huston, P. Nadolsky, J. Pumplin, C. Schmidt,D. Stump, and C. P. Yuan, New parton distribution functions from a global analysis ofquantum chromodynamics , Phys. Rev.
D93 (2016), no. 3 033006, [ arXiv:1506.07443 ].[7] L. A. Harland-Lang, A. D. Martin, P. Motylinski, and R. S. Thorne,
Parton distributionsin the LHC era: MMHT 2014 PDFs , Eur. Phys. J.
C75 (2015) 204, [ arXiv:1412.3989 ].[8] S. Alekhin, J. Blumlein, S. Moch, and R. Placakyte,
Parton Distribution Functions, α s and Heavy-Quark Masses for LHC Run II , arXiv:1701.05838 .[9] ZEUS, H1
Collaboration, H. Abramowicz et al.,
Combination of measurements ofinclusive deep inelastic e ± p scattering cross sections and QCD analysis of HERA data , Eur. Phys. J.
C75 (2015), no. 12 580, [ arXiv:1506.06042 ].[10] P. Jimenez-Delgado and E. Reya,
Delineating parton distributions and the strongcoupling , Phys.Rev.
D89 (2014), no. 7 074049, [ arXiv:1403.1852 ].[11] A. Accardi, L. T. Brady, W. Melnitchouk, J. F. Owens, and N. Sato,
Constraints onlarge- x parton distributions from new weak boson production and deep-inelastic scatteringdata , Phys. Rev.
D93 (2016), no. 11 114017, [ arXiv:1602.03154 ].[12] J. Butterworth et al.,
PDF4LHC recommendations for LHC Run II , J. Phys.
G43 (2016) 023001, [ arXiv:1510.03865 ].[13] D0 Collaboration, V. M. Abazov et al.,
Measurement of the muon charge asymmetry in p ¯ p → W+X → µν + X events at √ s =1.96 TeV , Phys.Rev.
D88 (2013) 091102,[ arXiv:1309.2591 ].[14] D0 Collaboration, V. M. Abazov et al.,
Measurement of the electron charge asymmetry in p ¯ p → W + X → eν + X decays in p ¯ p collisions at √ s = 1 . TeV , Phys. Rev.
D91 (2015), no. 3 032007, [ arXiv:1412.2862 ]. [Erratum: Phys. Rev.D91,no.7,079901(2015)].[15] M. Czakon, D. Heymes, and A. Mitov,
High-precision differential predictions fortop-quark pairs at the LHC , Phys. Rev. Lett. (2016), no. 8 082003,[ arXiv:1511.00549 ].[16] M. Czakon, D. Heymes, and A. Mitov,
Dynamical scales for multi-TeV top-pairproduction at the LHC , JHEP (2017) 071, [ arXiv:1606.03350 ].9317] R. Boughezal, C. Focke, X. Liu, and F. Petriello, W -boson production in association witha jet at next-to-next-to-leading order in perturbative QCD , Phys. Rev. Lett. (2015),no. 6 062002, [ arXiv:1504.02131 ].[18] A. Gehrmann-De Ridder, T. Gehrmann, E. W. N. Glover, A. Huss, and T. A. Morgan,
Precise QCD predictions for the production of a Z boson in association with a hadronicjet , Phys. Rev. Lett. (2016), no. 2 022001, [ arXiv:1507.02850 ].[19] J. Currie, A. Gehrmann-De Ridder, E. Glover, and J. Pires,
NNLO QCD corrections tojet production at hadron colliders from gluon scattering , JHEP (2014) 110,[ arXiv:1310.3993 ].[20] J. Currie, E. W. N. Glover, and J. Pires,
NNLO QCD predictions for single jet inclusiveproduction at the LHC , Phys. Rev. Lett. (2017), no. 7 072002, [ arXiv:1611.01460 ].[21] R. D. Ball, V. Bertone, M. Bonvini, S. Forte, P. Groth Merrild, J. Rojo, and L. Rottoli,
Intrinsic charm in a matched general-mass scheme , Phys. Lett.
B754 (2016) 49–58,[ arXiv:1510.00009 ].[22] R. D. Ball, M. Bonvini, and L. Rottoli,
Charm in Deep-Inelastic Scattering , JHEP (2015) 122, [ arXiv:1510.02491 ].[23] NNPDF
Collaboration, R. D. Ball, V. Bertone, M. Bonvini, S. Carrazza, S. Forte,A. Guffanti, N. P. Hartland, J. Rojo, and L. Rottoli,
A Determination of the CharmContent of the Proton , Eur. Phys. J.
C76 (2016), no. 11 647, [ arXiv:1605.06515 ].[24] S. Carrazza, S. Forte, Z. Kassabov, J. I. Latorre, and J. Rojo,
An Unbiased HessianRepresentation for Monte Carlo PDFs , Eur. Phys. J.
C75 (2015), no. 8 369,[ arXiv:1505.06736 ].[25] S. Carrazza, J. I. Latorre, J. Rojo, and G. Watt,
A compression algorithm for thecombination of PDF sets , Eur. Phys. J.
C75 (2015) 474, [ arXiv:1504.06469 ].[26] S. Carrazza, S. Forte, Z. Kassabov, and J. Rojo,
Specialized minimal PDFs for optimizedLHC calculations , Eur. Phys. J.
C76 (2016), no. 4 205, [ arXiv:1602.00005 ].[27] S. Carrazza and Z. Kassabov,
Parton Distribution Functions at LHC and the SMPDFweb-based application , PoS
PP@LHC2016 (2016) 020, [ arXiv:1606.09248 ].[28]
New Muon
Collaboration, M. Arneodo et al.,
Accurate measurement of F d /F p and R d − R p , Nucl. Phys.
B487 (1997) 3–26, [ hep-ex/9611022 ].[29]
New Muon
Collaboration, M. Arneodo et al.,
Measurement of the proton and deuteronstructure functions, F p and F d , and of the ratio σ L /σ T , Nucl. Phys.
B483 (1997) 3–43,[ hep-ph/9610231 ].[30]
BCDMS
Collaboration, A. C. Benvenuti et al.,
A high statistics measurement of theproton structure functions f ( x, q ) and r from deep inelastic muon scattering at high q , Phys. Lett.
B223 (1989) 485.[31]
BCDMS
Collaboration, A. C. Benvenuti et al.,
A high statistics measurement of thedeuteron structure functions f ( x, q ) and r from deep inelastic muon scattering at high q , Phys. Lett.
B237 (1990) 592. 9432] L. W. Whitlow, E. M. Riordan, S. Dasu, S. Rock, and A. Bodek,
Precise measurementsof the proton and deuteron structure functions from a global analysis of the SLAC deepinelastic electron scattering cross-sections , Phys. Lett.
B282 (1992) 475–482.[33]
H1 and ZEUS
Collaboration, F. Aaron et al.,
Combined Measurement and QCDAnalysis of the Inclusive e ± p Scattering Cross Sections at HERA , JHEP (2010)109, [ arXiv:0911.0884 ].[34] H1 Collaboration, F. Aaron et al.,
Inclusive Deep Inelastic Scattering at High Q withLongitudinally Polarised Lepton Beams at HERA , JHEP (2012) 061,[ arXiv:1206.7007 ].[35] H1 Collaboration, F. Aaron et al.,
Measurement of the Inclusive e ± p Scattering CrossSection at High Inelasticity y and of the Structure Function F L , Eur.Phys.J.
C71 (2011)1579, [ arXiv:1012.4355 ].[36]
ZEUS
Collaboration, A. Cooper Sarkar,
Measurement of high- Q neutral current deepinelastic e+p scattering cross sections with a longitudinally polarised positron beam atHERA , arXiv:1208.6138 .[37] ZEUS
Collaboration, H. Abramowicz et al.,
Measurement of high- Q charged currentdeep inelastic scattering cross sections with a longitudinally polarised positron beam atHERA , Eur.Phys.J.
C70 (2010) 945–963, [ arXiv:1008.3493 ].[38]
H1 , ZEUS
Collaboration, H. Abramowicz et al.,
Combination and QCD Analysis ofCharm Production Cross Section Measurements in Deep-Inelastic ep Scattering atHERA , Eur.Phys.J.
C73 (2013) 2311, [ arXiv:1211.1182 ].[39]
CHORUS
Collaboration, G. Onengut et al.,
Measurement of nucleon structurefunctions in neutrino scattering , Phys. Lett.
B632 (2006) 65–75.[40]
NuTeV
Collaboration, M. Goncharov et al.,
Precise measurement of dimuon productioncross-sections in ν µ Fe and ¯ ν µ Fe deep inelastic scattering at the Tevatron , Phys. Rev.
D64 (2001) 112006, [ hep-ex/0102049 ].[41] D. A. Mason,
Measurement of the strange - antistrange asymmetry at NLO in QCD fromNuTeV dimuon data , . FERMILAB-THESIS-2006-01.[42]
CDF
Collaboration, T. A. Aaltonen et al.,
Measurement of dσ/dy of Drell-Yan e + e − pairs in the Z Mass Region from p ¯ p Collisions at √ s = 1 . TeV , Phys. Lett.
B692 (2010) 232–239, [ arXiv:0908.3914 ].[43] D0 Collaboration, V. M. Abazov et al.,
Measurement of the shape of the boson rapiditydistribution for p ¯ p → Z/γ ∗ → e + e − + X events produced at √ s =1.96-TeV , Phys. Rev.
D76 (2007) 012003, [ hep-ex/0702025 ].[44]
CDF
Collaboration, T. Aaltonen et al.,
Measurement of the Inclusive Jet Cross Sectionat the Fermilab Tevatron p-pbar Collider Using a Cone-Based Jet Algorithm , Phys. Rev.
D78 (2008) 052006, [ arXiv:0807.2204 ].[45] G. Moreno et al.,
Dimuon production in proton - copper collisions at √ s = 38.8-GeV , Phys. Rev.
D43 (1991) 2815–2836.[46]
NuSea
Collaboration, J. C. Webb et al.,
Absolute Drell-Yan dimuon cross sections in800-GeV/c p p and p d collisions , hep-ex/0302019 .9547] J. C. Webb, Measurement of continuum dimuon production in 800-GeV/c proton nucleoncollisions , hep-ex/0301031 .[48] FNAL E866/NuSea
Collaboration, R. S. Towell et al.,
Improved measurement of theanti-d/anti-u asymmetry in the nucleon sea , Phys. Rev.
D64 (2001) 052002,[ hep-ex/0103030 ].[49]
ATLAS
Collaboration, G. Aad et al.,
Measurement of the inclusive W ± and Z/γ ∗ crosssections in the electron and muon decay channels in pp collisions at √ s = 7 TeV with theATLAS detector , Phys.Rev.
D85 (2012) 072004, [ arXiv:1109.5141 ].[50]
ATLAS
Collaboration, G. Aad et al.,
Measurement of the high-mass Drell–Yandifferential cross-section in pp collisions at √ s =7 TeV with the ATLAS detector , Phys.Lett.
B725 (2013) 223, [ arXiv:1305.4192 ].[51]
ATLAS
Collaboration, G. Aad et al.,
Measurement of the Transverse MomentumDistribution of W Bosons in pp Collisions at √ s = 7 TeV with the ATLAS Detector , Phys.Rev.
D85 (2012) 012005, [ arXiv:1108.6308 ].[52]
CMS
Collaboration, S. Chatrchyan et al.,
Measurement of the electron chargeasymmetry in inclusive W production in pp collisions at √ s = 7 TeV , Phys.Rev.Lett. (2012) 111806, [ arXiv:1206.2598 ].[53]
CMS
Collaboration, S. Chatrchyan et al.,
Measurement of the muon charge asymmetryin inclusive pp to WX production at √ s = 7 TeV and an improved determination of lightparton distribution functions , Phys.Rev.
D90 (2014) 032004, [ arXiv:1312.6283 ].[54]
CMS
Collaboration, S. Chatrchyan et al.,
Measurement of the differential anddouble-differential Drell-Yan cross sections in proton-proton collisions at √ s = , JHEP (2013) 030, [ arXiv:1310.7291 ].[55]
LHCb
Collaboration, R. Aaij et al.,
Inclusive W and Z production in the forward regionat √ s = 7 TeV , JHEP (2012) 058, [ arXiv:1204.1620 ].[56]
LHCb
Collaboration, R. Aaij et al.,
Measurement of the cross-section for Z → e + e − production in pp collisions at √ s = 7 TeV , JHEP (2013) 106, [ arXiv:1212.4620 ].[57]
ATLAS
Collaboration, G. Aad et al.,
Measurement of inclusive jet and dijet productionin pp collisions at √ s = 7 TeV using the ATLAS detector , Phys. Rev.
D86 (2012)014022, [ arXiv:1112.6297 ].[58]
ATLAS
Collaboration, G. Aad et al.,
Measurement of the inclusive jet cross section inpp collisions at √ s =2.76 TeV and comparison to the inclusive jet cross section at √ s =7TeV using the ATLAS detector , Eur.Phys.J.
C73 (2013) 2509, [ arXiv:1304.4739 ].[59]
CMS
Collaboration, S. Chatrchyan et al.,
Measurements of differential jet cross sectionsin proton-proton collisions at √ s = 7 TeV with the CMS detector , Phys.Rev.
D87 (2013)112002, [ arXiv:1212.6660 ].[60]
CMS
Collaboration, S. Chatrchyan et al.,
Measurement of associated W + charmproduction in pp collisions at √ s = 7 TeV , JHEP (2014) 013, [ arXiv:1310.1138 ].[61] ATLAS
Collaboration, G. Aad et al.,
Measurement of the cross section for top-quarkpair production in pp collisions at √ s = 7 TeV with the ATLAS detector using finalstates with two high-pt leptons , JHEP (2012) 059, [ arXiv:1202.4892 ].9662]
ATLAS
Collaboration, G. Aad et al.,
Measurement of the t ¯ t production cross-section inpp collisions at sqrts = 7 TeV using kinematic information of lepton+jets events , ATLAS-CONF-2011-121, ATLAS-COM-CONF-2011-132 .[63]
ATLAS
Collaboration, G. Aad et al.,
Measurement of the t ¯ t production cross-section in pp collisions at √ s = 8 TeV using eµ events with b -tagged jets , ATLAS-CONF-2013-097,ATLAS-COM-CONF-2013-112 .[64]
CMS
Collaboration, S. Chatrchyan et al.,
Measurement of the t ¯ t production crosssection in the dilepton channel in pp collisions at √ s = 8 TeV , JHEP (2014) 024,[ arXiv:1312.7582 ].[65]
CMS
Collaboration, S. Chatrchyan et al.,
Measurement of the t ¯ t production crosssection in the dilepton channel in pp collisions at √ s = 7 TeV , JHEP (2012) 067,[ arXiv:1208.2671 ].[66]
CMS
Collaboration, S. Chatrchyan et al.,
Measurement of the t ¯ t production crosssection in pp collisions at √ s = 7 TeV with lepton + jets final states , Phys.Lett.
B720 (2013) 83–104, [ arXiv:1212.6682 ].[67] H1 Collaboration, F. D. Aaron et al.,
Measurement of the Charm and Beauty StructureFunctions using the H1 Vertex Detector at HERA , Eur. Phys. J.
C65 (2010) 89–109,[ arXiv:0907.2643 ].[68]
ZEUS
Collaboration, H. Abramowicz et al.,
Measurement of beauty and charmproduction in deep inelastic scattering at HERA and measurement of the beauty-quarkmass , JHEP (2014) 127, [ arXiv:1405.6915 ].[69] European Muon
Collaboration, J. J. Aubert et al.,
Production of charmed particles in250-GeV µ + - iron interactions , Nucl. Phys.
B213 (1983) 31–64.[70]
HERAFitter developers’ Team
Collaboration, S. Camarda et al.,
QCD analysis of W - and Z -boson production at Tevatron , Eur. Phys. J.
C75 (2015), no. 9 458,[ arXiv:1503.05221 ].[71]
ATLAS
Collaboration, G. Aad et al.,
Measurement of the transverse momentum and φ ∗ η distributions of DrellYan lepton pairs in protonproton collisions at √ s = 8 TeV with theATLAS detector , Eur. Phys. J.
C76 (2016), no. 5 291, [ arXiv:1512.02192 ].[72]
ATLAS
Collaboration, M. Aaboud et al.,
Precision measurement and interpretation ofinclusive W + , W − and Z/γ ∗ production cross sections with the ATLAS detector , arXiv:1612.03016 .[73] ATLAS
Collaboration, G. Aad et al.,
Measurements of top-quark pair differentialcross-sections in the lepton+jets channel in pp collisions at √ s = 8 TeV using theATLAS detector , Eur. Phys. J.
C76 (2016), no. 10 538, [ arXiv:1511.04716 ].[74]
ATLAS
Collaboration, G. Aad et al.,
Measurement of the t ¯ t production cross-sectionusing eµ events with b-tagged jets in pp collisions at √ s = 7 and 8 TeV with the ATLASdetector , Eur. Phys. J.
C74 (2014), no. 10 3109, [ arXiv:1406.5375 ]. [Addendum: Eur.Phys. J.C76,no.11,642(2016)].[75]
ATLAS
Collaboration, M. Aaboud et al.,
Measurement of the t ¯ t productioncross-section using eµ events with b-tagged jets in pp collisions at √ s =13 TeV with theATLAS detector , Phys. Lett.
B761 (2016) 136–157, [ arXiv:1606.02699 ].9776]
ATLAS
Collaboration, G. Aad et al.,
Measurement of the inclusive jet cross-section inproton-proton collisions at √ s = 7 TeV using 4.5 fb of data with the ATLAS detector , JHEP (2015) 153, [ arXiv:1410.8857 ]. [Erratum: JHEP09,141(2015)].[77] ATLAS
Collaboration, G. Aad et al.,
Measurement of the low-mass Drell-Yandifferential cross section at √ s = 7 TeV using the ATLAS detector , JHEP (2014)112, [ arXiv:1404.1212 ].[78] ATLAS
Collaboration, G. Aad et al.,
Measurement of the
Z/γ ∗ boson transversemomentum distribution in pp collisions at √ s = 7 TeV with the ATLAS detector , JHEP (2014) 145, [ arXiv:1406.3660 ].[79] CMS
Collaboration, V. Khachatryan et al.,
Measurement of the differential cross sectionand charge asymmetry for inclusive pp → W ± + X production at √ s = 8 TeV , Eur.Phys. J.
C76 (2016), no. 8 469, [ arXiv:1603.01803 ].[80]
CMS
Collaboration, V. Khachatryan et al.,
Measurement of the inclusive jet crosssection in pp collisions at √ s = 2 . TeV , Eur. Phys. J.
C76 (2016), no. 5 265,[ arXiv:1512.06212 ].[81]
CMS
Collaboration, V. Khachatryan et al.,
Measurement of the differential cross sectionfor top quark pair production in pp collisions at √ s = 8 TeV , Eur. Phys. J.
C75 (2015),no. 11 542, [ arXiv:1505.04480 ].[82]
CMS
Collaboration, V. Khachatryan et al.,
Measurement of the t-tbar production crosssection in the e-mu channel in proton-proton collisions at sqrt(s) = 7 and 8 TeV , JHEP (2016) 029, [ arXiv:1603.02303 ].[83] CMS
Collaboration, V. Khachatryan et al.,
Measurement of the Z boson differentialcross section in transverse momentum and rapidity in protonproton collisions at 8 TeV , Phys. Lett.
B749 (2015) 187–209, [ arXiv:1504.03511 ].[84]
CMS
Collaboration, V. Khachatryan et al.,
Measurements of differential anddouble-differential Drell-Yan cross sections in proton-proton collisions at 8 TeV , Eur.Phys. J.
C75 (2015), no. 4 147, [ arXiv:1412.1115 ].[85]
LHCb
Collaboration, R. Aaij et al.,
Measurement of the forward Z boson productioncross-section in pp collisions at √ s = 7 TeV , JHEP (2015) 039, [ arXiv:1505.07024 ].[86] LHCb
Collaboration, R. Aaij et al.,
Measurement of forward W and Z boson productionin pp collisions at √ s = 8 TeV , JHEP (2016) 155, [ arXiv:1511.08039 ].[87] CDF - Run II
Collaboration, A. Abulencia et al.,
Measurement of the Inclusive JetCross Section using the k T algorithm in pp Collisions at √ s =1.96 TeV with the CDF IIDetector , Phys. Rev.
D75 (2007) 092006, [ hep-ex/0701051 ].[88]
CMS
Collaboration, C. Collaboration,
Measurement of the top quark pair productioncross section using e µ events in proton-proton collisions at √ s = 13 TeV with the CMSdetector , .[89] L. A. Harland-Lang and R. S. Thorne, “Private communication.” 2017.[90] R. S. Thorne, L. A. Harland-Lang, A. D. Martin, and P. Motylinski, The Effect of FinalHERA inclusive Cross Section Data MMHT2014 PDFs , in
Proceedings, 2015 EuropeanPhysical Society Conference on High Energy Physics (EPS-HEP 2015) , 2015. arXiv:1508.06621 . 9891] T.-J. Hou, S. Dulat, J. Gao, M. Guzzi, J. Huston, P. Nadolsky, J. Pumplin, C. Schmidt,D. Stump, and C. P. Yuan,
CTEQ-TEA parton distribution functions and HERA Run Iand II combined data , Phys. Rev.
D95 (2017), no. 3 034003, [ arXiv:1609.07968 ].[92] L. A. Harland-Lang, A. D. Martin, P. Motylinski, and R. S. Thorne,
The impact of thefinal HERA combined data on PDFs obtained from a global fit , Eur. Phys. J.
C76 (2016), no. 4 186, [ arXiv:1601.03413 ].[93] J. Rojo,
Progress in the NNPDF global analysis and the impact of the legacy HERAcombination , in
Proceedings, 2015 European Physical Society Conference on High EnergyPhysics (EPS-HEP 2015) , 2015. arXiv:1508.07731 .[94] L. A. Harland-Lang, A. D. Martin, P. Motylinski, and R. S. Thorne,
Charm and beautyquark masses in the MMHT2014 global PDF analysis , Eur. Phys. J.
C76 (2016), no. 110, [ arXiv:1510.02332 ].[95] L. Rottoli,
Fitting EMC structure functions with intrinsic charm , PoS
DIS2016 (2016)032, [ arXiv:1606.09289 ].[96]
Particle Data Group
Collaboration, C. Patrignani et al.,
Review of Particle Physics , Chin. Phys.
C40 (2016), no. 10 100001.[97]
The NNPDF
Collaboration, R. D. Ball et al.,
A determination of parton distributionswith faithful uncertainty estimation , Nucl. Phys.
B809 (2009) 1–63, [ arXiv:0808.1231 ].[98]
The NNPDF
Collaboration, R. D. Ball et al.,
A first unbiased global NLOdetermination of parton distributions and their uncertainties , Nucl. Phys.
B838 (2010)136, [ arXiv:1002.4407 ].[99] V. Bertone, S. Carrazza, and J. Rojo,
APFEL: A PDF Evolution Library with QEDcorrections , Comput.Phys.Commun. (2014) 1647, [ arXiv:1310.1394 ].[100] S. Forte, E. Laenen, P. Nason, and J. Rojo,
Heavy quarks in deep-inelastic scattering , Nucl. Phys.
B834 (2010) 116–162, [ arXiv:1001.2312 ].[101] S. Kretzer and I. Schienbein,
Heavy quark initiated contributions to deep inelasticstructure functions , Phys. Rev.
D58 (1998) 094035, [ hep-ph/9805233 ].[102] M. Bonvini, S. Marzani, J. Rojo, L. Rottoli, M. Ubiali, R. D. Ball, V. Bertone,S. Carrazza, and N. P. Hartland,
Parton distributions with threshold resummation , JHEP (2015) 191, [ arXiv:1507.01006 ].[103] C. Anastasiou, L. J. Dixon, K. Melnikov, and F. Petriello, High precision QCD at hadroncolliders: Electroweak gauge boson rapidity distributions at NNLO , Phys. Rev.
D69 (2004) 094008, [ hep-ph/0312266 ].[104] J. Rojo,
Constraints on parton distributions and the strong coupling from LHC jet data , Int. J. Mod. Phys.
A30 (2015) 1546005, [ arXiv:1410.7728 ].[105] Z. Nagy,
Three jet cross-sections in hadron hadron collisions at next-to-leading order , Phys.Rev.Lett. (2002) 122003, [ hep-ph/0110315 ].[106] T. Carli et al., A posteriori inclusion of parton density functions in NLO QCD final-statecalculations at hadron colliders: The APPLGRID Project , Eur.Phys.J.
C66 (2010) 503,[ arXiv:0911.2985 ]. 99107] S. Carrazza and J. Pires,
Perturbative QCD description of jet data from LHC Run-I andTevatron Run-II , JHEP (2014) 145, [ arXiv:1407.7031 ].[108] J. Currie, E. W. N. Glover, A. Gehrmann-De Ridder, T. Gehrmann, A. Huss, andJ. Pires, Single jet inclusive production for the individual jet p T scale choice at the LHC ,in , 2017. arXiv:1704.00923 .[109] G. P. Salam and J. Rojo, A Higher Order Perturbative Parton Evolution Toolkit(HOPPET) , Comput. Phys. Commun. (2009) 120–156, [ arXiv:0804.3755 ].[110] J. Currie, “Private communication.”.[111] T. Gleisberg et al.,
Event generation with SHERPA 1.1 , JHEP (2009) 007,[ arXiv:0811.4622 ].[112] L. Del Debbio, N. P. Hartland, and S. Schumann, MCgrid: projecting cross sectioncalculations on grids , Comput.Phys.Commun. (2014) 2115–2126,[ arXiv:1312.4460 ].[113] R. Gavin, Y. Li, F. Petriello, and S. Quackenbush,
FEWZ 2.0: A code for hadronic Zproduction at next-to-next-to-leading order , Comput. Phys. Commun. (2011)2388–2403, [ arXiv:1011.3540 ].[114] R. Gavin, Y. Li, F. Petriello, and S. Quackenbush,
W Physics at the LHC with FEWZ2.1 , Comput.Phys.Commun. (2013) 208–214, [ arXiv:1201.5896 ].[115] Y. Li and F. Petriello,
Combining QCD and electroweak corrections to dileptonproduction in FEWZ , Phys.Rev.
D86 (2012) 094034, [ arXiv:1208.5967 ].[116]
Mcfm , http://mcfm.fnal.gov .[117] S. Alekhin et al., HERAFitter , Eur. Phys. J.
C75 (2015), no. 7 304, [ arXiv:1410.4412 ].[118] A. Gehrmann-De Ridder, T. Gehrmann, E. W. N. Glover, A. Huss, and T. A. Morgan,
The NNLO QCD corrections to Z boson production at large transverse momentum , JHEP (2016) 133, [ arXiv:1605.04295 ].[119] R. Boughezal, J. M. Campbell, R. K. Ellis, C. Focke, W. T. Giele, X. Liu, andF. Petriello, Z-boson production in association with a jet at next-to-next-to-leading orderin perturbative QCD , Phys. Rev. Lett. (2016), no. 15 152001, [ arXiv:1512.01291 ].[120] R. Boughezal, X. Liu, and F. Petriello,
Phenomenology of the Z-boson plus jet process atNNLO , Phys. Rev.
D94 (2016), no. 7 074015, [ arXiv:1602.08140 ].[121] R. Boughezal, A. Guffanti, F. Petriello, and M. Ubiali,
The impact of the LHC Z-bosontransverse momentum data on PDF determinations , arXiv:1705.00343 .[122] CMS
Collaboration, S. Chatrchyan et al.,
Measurement of the Rapidity and TransverseMomentum Distributions of Z Bosons in pp Collisions at √ s = 7 TeV , Phys. Rev.
D85 (2012) 032002, [ arXiv:1110.4973 ].[123] S. Carrazza,
Modeling NNLO jet corrections with neural networks , in , 2017. arXiv:1704.00471 .100124] M. Czakon, N. P. Hartland, A. Mitov, E. R. Nocera, and J. Rojo,
Pinning down thelarge-x gluon with NNLO top-quark pair differential distributions , JHEP (2017) 044,[ arXiv:1611.08609 ].[125] M. Czakon, D. Heymes, and A. Mitov, Bump hunting in LHC t ¯ t events , Phys. Rev.
D94 (2016), no. 11 114033, [ arXiv:1608.00765 ].[126] A. Buckley, J. Butterworth, L. Lonnblad, D. Grellscheid, H. Hoeth, et al.,
Rivet usermanual , Comput.Phys.Commun. (2013) 2803–2819, [ arXiv:1003.0694 ].[127] F. Cascioli, P. Maierhofer, and S. Pozzorini,
Scattering Amplitudes with Open Loops , Phys. Rev. Lett. (2012) 111601, [ arXiv:1111.5206 ].[128]
Particle Data Group
Collaboration, K. Olive et al.,
Review of Particle Physics , Chin.Phys.
C38 (2014) 090001.[129] M. Czakon and A. Mitov,
Top++: A Program for the Calculation of the Top-PairCross-Section at Hadron Colliders , Comput. Phys. Commun. (2014) 2930,[ arXiv:1112.5675 ].[130] R. D. Ball,
Charm Production: Pole Mass or Running Mass? , AIP Conf. Proc. (2017), no. 1 030002, [ arXiv:1612.03790 ].[131]
LHC Higgs Cross Section Working Group
Collaboration, D. de Florian et al.,
Handbook of LHC Higgs Cross Sections: 4. Deciphering the Nature of the Higgs Sector , arXiv:1610.07922 .[132] The NNPDF
Collaboration, R. D. Ball et al.,
Impact of Heavy Quark Masses onParton Distributions and LHC Phenomenology , Nucl. Phys.
B849 (2011) 296,[ arXiv:1101.1300 ].[133] R. D. Ball et al.,
Parton distributions with LHC data , Nucl.Phys.
B867 (2013) 244,[ arXiv:1207.1303 ].[134] R. D. Ball and S. Forte,
Double asymptotic scaling at hera , Phys. Lett.
B335 (1994)77–86, [ hep-ph/9405320 ].[135]
ATLAS
Collaboration, G. Aad et al.,
Determination of the strange quark density of theproton from ATLAS measurements of the
W, Z cross sections , Phys.Rev.Lett. (2012)[ arXiv:1203.4051 ].[136] S. Dulat et al.,
Intrinsic Charm Parton Distribution Functions from CTEQ-TEA GlobalAnalysis , Phys.Rev.
D89 (2014), no. 7 073004, [ arXiv:1309.0025 ].[137] P. Jimenez-Delgado, T. Hobbs, J. Londergan, and W. Melnitchouk,
New limits onintrinsic charm in the nucleon from global analysis of parton distributions , Phys.Rev.Lett. (2015), no. 8 082002, [ arXiv:1408.1708 ].[138] M. L. Mangano et al.,
Physics at a 100 TeV pp collider: Standard Model processes , arXiv:1607.01831 .[139] ATLAS
Collaboration, G. Aad et al.,
Measurement of W ± and Z -boson productioncross sections in pp collisions at √ s = 13 TeV with the ATLAS detector , Phys. Lett.
B759 (2016) 601–621, [ arXiv:1603.09222 ].101140]
CMS Collaboration
Collaboration,
Measurement of inclusive W and Z bosonproduction cross sections in pp collisions at sqrt(s)=13 TeV , .[141] C. Carloni Calame, G. Montagna, O. Nicrosini, and A. Vicini,
Precision electroweakcalculation of the production of a high transverse-momentum lepton pair at hadroncolliders , JHEP (2007) 109, [ arXiv:0710.1722 ].[142] F. Bishara, R. Contino, and J. Rojo,
Higgs pair production in vector-boson fusion at theLHC and beyond , arXiv:1611.03860 .[143] J. K. Behr, D. Bortoletto, J. A. Frost, N. P. Hartland, C. Issever, and J. Rojo, BoostingHiggs pair production in the b ¯ bb ¯ b final state with multivariate techniques , Eur. Phys. J.
C76 (2016), no. 7 386, [ arXiv:1512.08928 ].[144] R. D. Ball, M. Bonvini, S. Forte, S. Marzani, and G. Ridolfi,
Higgs production in gluonfusion beyond NNLO , Nucl.Phys.
B874 (2013) 746, [ arXiv:1303.3590 ].[145] M. Bonvini, R. D. Ball, S. Forte, S. Marzani, and G. Ridolfi,
Updated Higgs cross sectionat approximate N LO , J. Phys.
G41 (2014) 095002, [ arXiv:1404.3204 ].[146] M. Bonvini, S. Marzani, C. Muselli, and L. Rottoli,
On the Higgs cross section atN LO+N LL and its uncertainty , JHEP (2016) 105, [ arXiv:1603.08000 ].[147] J. Alwall, R. Frederix, S. Frixione, V. Hirschi, F. Maltoni, et al., The automatedcomputation of tree-level and next-to-leading order differential cross sections, and theirmatching to parton shower simulations , JHEP (2014) 079, [ arXiv:1405.0301 ].[148] O. Brein, R. V. Harlander, and T. J. E. Zirke, vh@nnlo - Higgs Strahlung at hadroncolliders , Comput. Phys. Commun. (2013) 998–1003, [ arXiv:1210.5347 ].[149] F. A. Dreyer and A. Karlberg,
Vector-Boson Fusion Higgs Production at Three Loops inQCD , Phys. Rev. Lett. (2016), no. 7 072001, [ arXiv:1606.00840 ].[150] M. Cacciari, F. A. Dreyer, A. Karlberg, G. P. Salam, and G. Zanderighi,
FullyDifferential Vector-Boson-Fusion Higgs Production at Next-to-Next-to-Leading Order , Phys. Rev. Lett. (2015), no. 8 082002, [ arXiv:1506.02660 ].[151] A. Buckley, J. Ferrando, S. Lloyd, K. Nordstrm, B. Page, et al.,
LHAPDF6: partondensity access in the LHC precision era , Eur.Phys.J.
C75 (2015) 132,[ arXiv:1412.7420 ].[152] S. Lionetti et al.,
Precision determination of α s using an unbiased global NLO parton set , Phys. Lett.
B701 (2011) 346–352, [ arXiv:1103.2369 ].[153] R. D. Ball, V. Bertone, L. Del Debbio, S. Forte, A. Guffanti, et al.,
Precision NNLOdetermination of α s ( M Z ) using an unbiased global parton set , Phys.Lett.
B707 (2012)66–71, [ arXiv:1110.2483 ].[154]
NNPDF
Collaboration, R. D. Ball et al.,
Parton distributions with QED corrections , Nucl.Phys.
B877 (2013) 290–320, [ arXiv:1308.0598 ].[155] V. Bertone and S. Carrazza,
Combining NNPDF3.0 and NNPDF2.3QED through theAPFEL evolution code , PoS
DIS2016 (2016) 031, [ arXiv:1606.07130 ].[156] A. Manohar, P. Nason, G. P. Salam, and G. Zanderighi,
How bright is the proton? Aprecise determination of the photon parton distribution function , Phys. Rev. Lett. (2016), no. 24 242002, [ arXiv:1607.04266 ].102157] G. Altarelli, R. D. Ball, and S. Forte,
Perturbatively stable resummed small x evolutionkernels , Nucl. Phys.
B742 (2006) 1–40, [ hep-ph/0512237 ].[158] R. D. Ball,
Resummation of Hadroproduction Cross-sections at High Energy , Nucl. Phys.
B796 (2008) 137–183, [ arXiv:0708.1277 ].[159] G. Altarelli, R. D. Ball, and S. Forte,
Small x Resummation with Quarks: Deep-InelasticScattering , Nucl. Phys.
B799 (2008) 199–240, [ arXiv:0802.0032 ].[160] M. Bonvini, S. Marzani, and T. Peraro,
Small- x resummation from HELL , Eur. Phys. J.
C76 (2016), no. 11 597, [ arXiv:1607.02153 ].[161] F. Caola, S. Forte, and J. Rojo,
HERA data and DGLAP evolution: Theory andphenomenology , Nucl.Phys.
A854 (2011) 32–44, [ arXiv:1007.5405 ].[162] F. Caola, S. Forte, and J. Rojo,
Deviations from NLO QCD evolution in inclusive HERAdata , Phys. Lett.
B686 (2010) 127–135, [ arXiv:0910.3143 ].[163]
ATLAS
Collaboration, M. Aaboud et al.,
Measurement of the cross section for inclusiveisolated-photon production in pp collisions at √ s = 13 TeV using the ATLAS detector , arXiv:1701.06882 .[164] J. M. Campbell, R. K. Ellis, and C. Williams, Direct photon production atnext-to-next-to-leading order , arXiv:1612.04333 .[165] D. d’Enterria and J. Rojo, Quantitative constraints on the gluon distribution function inthe proton from collider isolated-photon data , Nucl.Phys.
B860 (2012) 311–338,[ arXiv:1202.1762 ].[166] M. Brucherseifer, F. Caola, and K. Melnikov,
On the NNLO QCD corrections tosingle-top production at the LHC , Phys. Lett.
B736 (2014) 58–63, [ arXiv:1404.7116 ].[167]
ATLAS
Collaboration, G. Aad et al.,
Measurement of the production of a W boson inassociation with a charm quark in pp collisions at √ s = , JHEP (2014) 068, [ arXiv:1402.6263 ].[168]
PROSA
Collaboration, O. Zenaiev et al.,
Impact of heavy-flavour production crosssections measured by the LHCb experiment on parton distribution functions at low x , Eur. Phys. J.
C75 (2015), no. 8 396, [ arXiv:1503.04581 ].[169] R. Gauld, J. Rojo, L. Rottoli, and J. Talbert,
Charm production in the forward region:constraints on the small-x gluon and backgrounds for neutrino astronomy , JHEP (2015) 009, [ arXiv:1506.08025 ].[170] R. Gauld and J. Rojo, Precision determination of the small- x gluon from charmproduction at LHCb , Phys. Rev. Lett. (2017), no. 7 072001, [ arXiv:1610.09373 ].[171] M. Botje,
QCDNUM: Fast QCD Evolution and Convolution , Comput.Phys.Commun. (2011) 490–532, [ arXiv:1005.1481 ].[172] S. Alekhin, J. Bluemlein, and S. Moch,
The ABM parton distributions tuned to LHCdata , Phys.Rev.
D89 (2014) 054028, [ arXiv:1310.3059arXiv:1310.3059