implications of w charge asymmetry measurements for pdf...
TRANSCRIPT
![Page 1: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/1.jpg)
Implications of W charge asymmetrymeasurements for PDF fits
Maria Ubiali(RWTH University Aachen)
in collaboration with:
F. Cerutti, J.I. Latorre (Barcelona U)R.D. Ball, L. Del Debbio (Edinburgh U)A. Guffanti, V. Bertone (Freiburg U)
S. Forte, J. Rojo (Milan U)
31st May 2011XXIII Rencontres de Blois - Particle Physics and Cosmology
NNPDFNNPDFcollaboration
![Page 2: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/2.jpg)
Summary
Introduction Parton Distribution Functions LHC and PDF uncertainties Reweighting techniques
The W lepton charge asymmetry data Effect of the Tevatron data on parton uncertainties Effect of the LHC data (CMS and ATLAS) on parton uncertainties A few more examples (Z rapidity distribution and W charm)
Conclusion and outlook
![Page 3: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/3.jpg)
LHC collisionsand PDFs uncertainty
A reliable understanding of PDFs and oftheir associated uncertainties plays acrucial role in relating theory toexperimental observation at the LHC.
G. Watt, PDF4LHC
P Px1P x2P
For many standard candleprocesses at the LHCtheoretical uncertaintyis dominated by PDFuncertainties
![Page 4: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/4.jpg)
LHC collisionsand PDFs uncertainty
Can LHC data help in thisdirection by providing more infoon PDFs and thereby reduce theiruncertainties & discriminatebetween models?
For many standard candleprocesses at the LHCtheoretical uncertaintyis dominated by PDFuncertainties
V. Radescu, DIS2011
![Page 5: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/5.jpg)
Is it possible to assess immediately theimpact of the new LHC data without refitting?
Bayesian reweightingand PDFs uncertainty
Bayesian Reweighting
Giele, Keller [hep-ph/9803393] NNPDF collaboration [ArXiv: 1012.0836] Can determine consistency, effect on PDF precision and shapes, impact on predictionson any other observable without refitting. Can be done quickly and easily by anybody (no need of PDF fitters): all you need todo is to compute the χ2 to the new data for each error set.
Conventional PDF fits: Whenever add new data, need todo full refitting, tune parametrizationand statistic treatment. Can be done only by PDF fittingcollaborations themselves.
Monte Carlo method (NNPDF): Always uses the same “infinite”parametrization (no tuning to data). Provides Monte Carlo ensemble in space ofPDFs: can add new data simply by updatingprobability.
![Page 6: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/6.jpg)
Generate a Monte Carloensemble in space of dataand project in space ofPDFs
“Infinite” neural networkparametrization for PDFsat the initial scale(300 pars versus typical~30 pars in polynomialfits)
Cross-validation methodto stop the fit dynamically
Full NLO and GM-VFNStheory
Probability density in thespace of PDFs
Bayesian reweightingThe NNPDF2.1 parton set
![Page 7: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/7.jpg)
Probability density in thespace of PDFs
Bayesian reweightingThe NNPDF2.1 parton set
![Page 8: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/8.jpg)
Probability density in thespace of PDFs (uniform sample)
Probability density of PDFsconditional on both “old” and“new” data (weighted sample)
χ2k of the kth replica to the NEW data
Compute Neff replicas: if Neff << N new dataare constraining or incompatible Compute P(α): if centered about 1 data arecompatible, if far from one needs tolerance α tofit them along with old data (incompatible)
Bayesian reweightingUpdate the (NN)PDFs
![Page 9: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/9.jpg)
The W lepton asymmetryTevatron Run II
CDF W charge asymmetry is fitted in NNPDF2.1 [ArXiv:0901.2169] D0 muon charge asymmetry in single pT
µ bin [ArXiv:0709.4254] D0 electron charge asymmetry combined andseparated pT
e bins :
[ArXiv: 0807.3367]
10-2 ≲ x ≲ 0.2
CT10 CT10w
![Page 10: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/10.jpg)
The W lepton asymmetryTevatron Run II
Inclusive bins in pTmuon
OK
Exclusive bins in pTel
KO
H. Schellman, DIS 2011Reweighting analysis[NNPDF, ArXiv:1012.0836] It is possible to include D0lepton asymmetry inclusive datain global analyses: no need ofproducing separate sets Issues with exclusive bins
![Page 11: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/11.jpg)
The W lepton asymmetryTevatron Run II
Reweighting analysis[NNPDF, ArXiv:1012.0836] It is possible to include D0lepton asymmetry inclusive datain global analyses: no need ofproducing separate sets Issues with exclusive bins
Reduction of the valence PDFs uncertainty and on the d/u ratio!
![Page 12: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/12.jpg)
The W lepton asymmetryLHC @ 7TeV measurements
ATLAS:W muon charge asymmetry (31pb-1)ArXiv: 1103.2929
CMS:W muon and electron charge asymmetry(36pb-1)ArXiv: 1103.3470
LHCb:Preliminary forward W muon chargeasymmetry (16.5pb-1)Not corrected for FSR radiation
ATL
AS, CM
SLH
Cb
LHCb
![Page 13: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/13.jpg)
The W lepton asymmetryLHC @ 7TeV predictions
NLO theoreticalpredictionsobtained withDYNNLO[ArXiv: 0903.2120]
![Page 14: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/14.jpg)
The W lepton asymmetryLHC @ 7TeV, ATLAS data
Neff / N ~ 90%
Data compatible with data includedin global PDF analysis Slight reduction of uncertainty atmedium-small x for light (anti)quark
![Page 15: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/15.jpg)
The W lepton asymmetryLHC @ 7TeV, CMS data
Neff / N ~ 60%
Data compatible with data included inglobal PDF analysis Significant reduction of uncertainty atmedium-small x for light (anti)quark
![Page 16: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/16.jpg)
The W lepton asymmetryLHC @ 7TeV, ATLAS & CMS data
ATLAS and CMS data can beadded at the same time, no clearsigns of tension CMS data are more constrainingthan ATLAS data. Inclusion of data in PDFs fitsreduces uncertainty of more than40% in the small-medium xregion for light (anti)quark PDFs
![Page 17: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/17.jpg)
The W lepton asymmetryCan we combine Tevatron and LHC data?
Neff / N ~ 25% YES data are very constrainingon PDFs but compatible with dataincluded in the global analyses
After reweighting thereweighted set fits very well bothTevatron (D0 muon and D0electron inclusive) and LHC(ATLAS and CMS) W leptonasymmetry data
The description of W asymmetryfrom CDF does not deteriorate
What about the PDFs central values and PDFs uncertainty?
![Page 18: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/18.jpg)
The W lepton asymmetryCan we combine Tevatron and LHC data?
Reduction ofuncertainty of light quarkand anti-quark PDFs insmall-medium x region isdriven by LHC data.
Shift and reduction ofuncertainty in light quarkPDFs driven by D0Tevatron data.
NNPDF2.2 parton setincluding these datais going to beavailable soon onLHAPDF
![Page 19: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/19.jpg)
The W lepton asymmetryCan we combine Tevatron and LHC data?
Reduction ofuncertainty of light quarkand anti-quark PDFs insmall-medium x region isdriven by LHC data.
Shift and reduction ofuncertainty in light quarkPDFs driven by D0Tevatron data.
NNPDF2.2 parton setincluding these datais going to beavailable soon onLHAPDF
![Page 20: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/20.jpg)
Future measurementsLHC pseudo-data analysis
W asymmetry + Z rapiditydistribution with 2% uncertainty
W-charm production, charm pseudo-rapidity distribution with 10% uncertainty
Generate pseudo-data fluctuating about predictions produced with NNPDF2.1within a given % (statistical) uncertainty Check the effect of future measurements on PDF shapes
antiup strange
![Page 21: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/21.jpg)
Conclusions and Outlook LHC data (and Tevatron) provide important constraint PDFs
• 40% in the small/medium-x region from ATLAS and CMS• 20% in the medium-x region from D0• NNPDF2.2 will include these info - available soon on LHAPDF!
As statistics increases uncertainty can be further reduced and precise LHCdata can discriminate among different PDF models
Is NNPDF2.1 HERA + Tevatron + LHC (without fixed-target data) the future?• Medium and large x gluon:
Prompt photon ✔Precision jets data ✔
• Light flavors at medium and small xLow-mass Drell-Yan ✔Z rapidity distributions ✔W asymmetries ✔ and polarized W ✔
• Strangeness and heavy flavorsWc for strangeness ✔Zc and γc for charm ✔Zb for bottom ✔
![Page 22: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/22.jpg)
LHC data (and Tevatron) provide important constraint PDFs• 40% in the small/medium-x region from ATLAS and CMS• 20% in the medium-x region from D0• NNPDF2.2 will include these info - available soon on LHAPDF!
As statistics increases uncertainty can be further reduced and precise LHCdata can discriminate among different PDF models
Is NNPDF2.1 HERA + Tevatron + LHC (without fixed-target data) the future?• Medium and large x gluon:
Prompt photon ✔Precision jets data ✔
• Light flavors at medium and small xLow-mass Drell-Yan ✔Z rapidity distributions ✔W asymmetries ✔ and polarized W ✔
• Strangeness and heavy flavorsWc for strangeness ✔Zc and γc for charm ✔Zb for bottom ✔
Conclusions and Outlook
THANKS FORTHANKS FOR YOUR ATTENTION !!YOUR ATTENTION !!
![Page 23: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/23.jpg)
BACKUP
![Page 24: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/24.jpg)
LHC collisionsand PDFs uncertainty
What data do constrain PDFs?
DIS neutral and chargedcurrent data (singlet-tripletseparation) + F2
c,F2b
Neutrino data, inclusive andcharm tagged (strange andanti-strange)
Drell-Yan data from fixedtarget (anti-up, anti-downseparation)
Vector boson production andasymmetries from Tevatron(up-down flavor asymmetries)
Inclusive jet data (gluon)
![Page 25: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/25.jpg)
Backup
![Page 26: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/26.jpg)
Backup
![Page 27: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/27.jpg)
Backup
![Page 28: Implications of W charge asymmetry measurements for PDF fitslss.fnal.gov/conf2/C110529/ubiali.pdf · “Infinite” neural network parametrization for PDFs at the initial scale (300](https://reader033.vdocuments.mx/reader033/viewer/2022050200/5f540f3e87d5885c756f2b47/html5/thumbnails/28.jpg)
Backup