NNLO PDF fits with top-quark pair differential distributions

NNLO PDF fits with !top-quark pair differential distributions

Juan Rojo!VU Amsterdam & Theory group, Nikhef!

!Based on arxiv:1611.aaaaa, !

with M. Czakon, N. P. Hartland, A. Mitov, and E. R. Nocera!!

with many thanks for assistance and discussions to M. Aldaya, F. Deliot, A. Gianmanco, !R. Gonzalez, A. Lister, A. Meyer, M. Owen, P. Silva, F. Spano, …!

!LHC Top Working Group meeting!

CERN, 21/11/2016

Juan Rojo LHC Top Working group meeting, 21/09/20161

2

!

Total cross-sections already used in many PDF fits, using differential measurements will only increase constraining power!

NNLO calculation for stable top quarks available, together with scale optimization, allows for the use of state-of-the art theory for top data in global PDF fits!

Available precise data from ATLAS and CMS at 8 TeV with full breakdown of statistical and systematic uncertainties!

Study interplay with inclusive jet production measurements (NNLO calculation very recently completed)

Why top-pair differential data in PDF fits?


Czakon, Mangano, Mitov, Rojo 13

ABM12 MMHT14

3

PDF fit settings


!

Based on the NNPDF3 global analysis framework!

Baseline fit based on the same dataset as in NNPDF3.0, except for:!

Separate H1 and ZEUS data from HERA-II replaced by final HERA combination!

Jet data excluded (since NNLO for inclusive jets only very recently available)!

NNLO theory, with αS(mZ) =0.118, mtop=173.3 GeV, mcharm =1.51 GeV, mbottom=4.92 GeV!

Optimized choice of factorization and renormalization scales

!

Fast NLO calculations based on Sherpa/MCgrid, supplemented by bin-by-bin NNLO/NLO K-factors!

All available sources of statistical and systematic correlated uncertainties accounted for

For ptT!

For yt,ytt, mtt!

Czakon,Heines, Mitov 16

NNPDF Collaboration 14-16

Currie, Glover, Pires 16

4

Top-pair differential distributions


!

Focus on the lepton+jets 8 TeV data from ATLAS and CMS (no technical limitations to include other differential measurements)!

ATLAS arXiv:1511.04716!CMS arXiv:1505.04480!

! Fit either absolute or

normalized distributions (in the latter case adding total cross-sections) and compare results!

Statistical correlations between distributions unknown: can only include one distribution per experiment in the PDF fit

5

NNLO corrections


!

For the top and top-pair rapidity, K-factors are roughly flat with size between 1.06 and 1.08!

For ptT, K-factor decreases from 1.1 at low ptT to 1 at high ptT!

For mtt , K-factor is al large as 1.12 above 1 TeV!

K-factors independent of the input PDFs used in their calculation, to good approximation

6

Sensitivity to the large-x gluon


! The correlation coefficient between the gluon PDF and each of the bins of the four kinematic

distributions determines the region of Bjorken-x where available data has sensitivity!

Large values, 𝜌 ≥ 0.8, of the correlation for 0.07 < x < 0.6, representing a significant improved over the coverage in x of inclusive cross-section measurements

𝜌 = 0.8

𝜌 = 0.8

7

NNLO theory confronts LHC data


8



!

Normalized distributions, of course, exhibit much smaller experimental uncertainties!

However, the size of the ATLAS and CMS uncertainties can differ by substantial amounts!

Specially in then normalized case, results not always consistent between the two experiments

9



!

At the qualitative level, NNPDF3.0, MMHT14, CT14 and HERAPDF2.0 reasonable agreement with the data (but need to compute 𝜒2/Ndat for a quantitative assessment)!

Larger differences are found for ABM12, both for absolute (systematic undershooting) as well as for normalized distributions

10



11



!

For mtt , exp uncertainties similar between abs and norm distributions at large masses!

In the highest bin, ATLAS uncertainty 4 times larger than the CMS one!

Again, for some bins the measurements from the two experiments are not really consistent

12



13

𝜒2/Ndat for the NNLO global fits


!

Entries in boldface indicate the datasets included in each specific fit

14



!

Poor 𝜒2/Ndat for some distributions improved when ATLAS and CMS data are fitted separately

Fitting separately!ATLAS and CMS data

1.1

2.6

1.8

4.1

15




1.1

2.6

1.8

4.1


0.8

0.6

1.7

0.9

!in a HERA-only fit

!

Further improvement of data/theory agreement when ATLAS and CMS data added separately on top of a HERA-only fit: discard possible internal inconsistencies or limitations of NNLO QCD theory

16

Impact on the large-x gluon


Absolute Normalized

17

Which top-quark differential data to fit?


Our recommendation for the 8 TeV lepton+jets differential distributions:

!

Fit normalized distributions, which exhibit somewhat enhanced constraining power!

Among these, select the distributions with which lead to larger reduction of PDF uncertainties!

Require good agreement between data and theory, 𝜒2/Ndat ~ 1, to avoid distorting the fit in the case of (still not understood) tensions between ATLAS and CMS!

Include one kinematic distribution from ATLAS and a different one from CMS, to achieve better kinematical coverage on the gluon PDF

!

Other possible choices, within the above guidelines, would lead to consistent results, since the pull on the large-x gluon is similar for all the ATLAS and CMS distributions (even when 𝜒2/Ndat » 1)

18

Impact on the large-x gluon


!

Significant reduction of PDF uncertainties in the gluon-gluon luminosity at high invariant masses!

For instance, for MX=2 TeV, the PDF uncertainties decrease from 13% to 5%!

Remarkably, the constraints from top differential data in the global fit are comparable to those from inclusive jets, despite coming from much fewer data points: Ndat =17 for top vs Ndat =470 for jets

MX=2 TeV

19

Impact on differential distributions


none of the LHC data showed in these plots has been used in the PDF fit

20

Impact on differential distributions


!

PDF uncertainties reduced by more than a factor two for mtt > 650 GeV!

Similar improvements for gluon-driven processes, either SM or BSM, at high masses!

Self-consistent program to use top data to provide improved theory predictions!

Our choice of fitted distributions (yt and ytt) reduce the risk of BSM contamination (kinematical suppression of heavy resonances) which can show up instead in the tails of the mtt and ptT distributions (whose PDF uncertainties are now greatly reduced)

Improved sensitivity to BSM physics !with top-quark final states

21 Juan Rojo LHC Top Working group meeting, 21/09/2016

Summary and outlook!

Differential distributions from top-quark pair production provide stringent constraints on the large-x gluon, comparable to those from inclusive jet production!

The possibility to fit some distributions (yt and ytt) and use the results to provide updated predictions for others (ptT and mtt) leads to a significant reduction of theory uncertainties with minimal risk of BSM contamination!

Some distributions exhibit a tension between the ATLAS and CMS measurements. The quality of the data description is significantly improved if the two experiments are fitted separately!

Further comparisons using other final states, 13 TeV data and particle-level distributions should shed more light on the origin of these tensions!

Differential top quark measurements will be an essential ingredient !for the next generation of global PDF fits!


Additional Material

NNLO PDF fits with top-quark pair differential distributions

Education

Transcript of NNLO PDF fits with top-quark pair differential distributions