PARTON DISTRIBUTIONS FOR LHC RUN II STEFANO FORTE UNIVERSITÀ DI MILANO & INFN

Transcription

1 PARTON DISTRIBUTIONS FOR LHC RUN II STEFANO FORTE UNIVERSITÀ DI MILANO & INFN LAL ORSAY, JULY 3, 5

2 PROLOGUE PDFS FOR LHC RUN I

3 WHY WORRY ABOUT PDFS? HIGGS PRODUCTION (J. Campbell, HCP) PDF UNCERTAINTY EITHER DOMINANT, OR VERY LARGE, OR BOTH... AND NOT ONLY FOR THE HIGGS! (W MASS DETERMINATION, NEW PHYSICS SEARCHES FOR HEAVY STATES,... )

4 σ X (s,m X ) = a,b PDFS FOR LHC RUN I x min dx dx f a/h (x )f b/h (x )ˆσ q aq b X ( ) x x s,mx LHC KINEMATICS NNPDF.3 Dataset 9 LHC parton kinematics ] [ GeV / p / M Q 7 6 T NMCPD NMC SLAC BCDMS HERAAV CHORUS FLH8 NTVDMN ZEUSH ZEUSFC HFC DYE886 DYE65 CDFWASY CDFZRAP DZRAP ATLASWZRAP DRCON CDFRKT ATLASR4JETS CMSWEASY LHCBW Q (GeV ) x, = (M/4 TeV) exp(±y) Q = M M = GeV M = GeV M = TeV HERA M = TeV y = fixed target x x MSTW8 CT NNPDF.3 HERAPDF.5 ABM/ JR9 HERA DIS FIXED TARGET DIS FIXED TARGET DY TEVATRON W +Z+JETS LHC W +Z+JETS some, approx.

5 RUN I PDF SETS: THE APPROACH METHODOLOGY STATISTICAL TREATMENT: CTEQ, MSTW HESSIAN WITH DYNAMICAL TOLERANCE; HERAPDF, STANDARD HESSIAN+PARM. ERROR ANALYSIS; GJR, HESSIAN WITH FIXED TOLERANCE; ABKM STANDARD HESSIAN; NNPDF MONTE CARLO (ALSO STUDIED BY HERAPDF, MSTW) PARTON PARAMETRIZATION: CTEQ, MSTW, HERAPDF x α ( x) β POLYNOMIALS; GJR: DITTO + VALENCELIKE ASSUMPTION; NNPDF NEURAL NETS; CHEBYSHEV POLYNOMIALS STUDIED BY HERAPDF, MSTW; COVARIANCE MATRIX, NORMALIZATION UNCERTAINTIES, OUTLIERS, THEORETICAL UNCERTAINTIES... THEORY α s VALUE: CTEQ: EXTERNAL PARAMETER, SEVERAL VALUES AVAILABLE; NNPDF: EXTERNAL PARAMETER, SEVERAL VALUES AVAILABLE, BEST FIT DETERMINED; MSTW: FITTED, BUT ALSO VARIABLE AS EXT.PARAMETER; ABKM: FITTED, VARIABLE AS EXT.PARAMETER (ONLY CENTRAL VALUE); GJR: FITTED, NOT VARIABLE AS EXT. PARAMETER; HEAVY QUARKS: CTEQ: GM VFN (SACOT χ SCHEME); MSTW: GM VFN (ACOT+TR SCHEME); NNPDF: GM VFN (FONLL SCHEME); ABKM: FFN (N f = 3,4 MATCHED WITH BMSN SCHEME); GJR: FFN (N f = 3) NUCLEAR CORRECTIONS, HIGHER TWISTS, KINEMATIC CUTS, INITIAL SCALE,... MSTW8 CT NNPDF.3 HERAPDF.5 ABM/ JR9 NO. OF PDFS STATISTICS HESS.+DT HESS.+DT MC HESS.+MODEL+PARM. HESS. HESS.+T PDF PARMS HEAVY QUARKS VFN TR VFN ACOT VFN FONLL VFN TR FFN FFN

6 ) B(lν) [pb] + σ(w W + LHC 8 TeV σ(w ) - VRAP NNLO - α S =.8 NNPDF.3 MSTW8 CT ABM HERAPDF.5 CMS LHC EW STANDARD CANDLES ) B(lν) [pb] - σ(w W - LHC 8 TeV σ(w ) - VRAP NNLO - α S =.8 NNPDF.3 MSTW8 CT ABM HERAPDF.5 CMS ) [pb] - l + ) B(l σ(z Z LHC 8 TeV σ(z ) - VRAP NNLO - α S =.8 NNPDF.3 MSTW8 CT ABM HERAPDF.5 CMS AGREEMENT/DISAGREEMENT BY DATA DRIVEN DIS ONLY FIT (HERAPDF) SAFE, BUT LARGE UNCERTAINTY FITS WITH SMALLER DATASETS PRONE TO TH. BIAS

7 PARTON LUMINOSITIES: QUARK SECTOR (q q) GLOBAL PDF SETS (ratio to NNPDF.3) OTHER PDF SETS (ratio to NNPDF.3).3 LHC 8 TeV - Ratio to NNPDF.3 NNLO - α s =.8.3 LHC 8 TeV - Ratio to NNPDF.3 NNLO - α s =.8 Quark - Antiquark Luminosity NNPDF.3 NNLO CT NNLO MSTW8 NNLO Quark - Antiquark Luminosity NNPDF.3 NNLO ABM NNLO HERAPDF.5 NNLO.8 M X 3.8 M X 3 CROSS SECTIONS REFLECT UNDERLYING LUMINOSITIES FEWER DATA LARGER UNCERTAINTIES (OR SYSTEMATIC BIAS) GLOBAL SETS: GOOD AGREEMENT IN THE REGION OF THE EW SCALE UNCERTAINTIES BLOW UP FOR LARGE MASS FINAL STATES

8 PARTON LUMINOSITIES: GLUON SECTOR GLOBAL PDF SETS (ratio to NNPDF.3) OTHER PDF SETS (ratio to NNPDF.3).3 LHC 8 TeV - Ratio to NNPDF.3 NNLO - α s =.8.3 LHC 8 TeV - Ratio to NNPDF.3 NNLO - α s =.8 Gluon - Gluon Luminosity NNPDF.3 NNLO CT NNLO MSTW8 NNLO Gluon - Gluon Luminosity NNPDF.3 NNLO ABM NNLO HERAPDF.5 NNLO M X 3.8 M X 3 FEWER DATA LARGER UNCERTAINTIES (OR SYSTEMATIC BIAS) GLOBAL SETS: NOT SO GOOD AGREEMENT IN THE REGION OF THE EW SCALE UNCERTAINTIES BLOW UP FOR LARGE MASS FINAL STATES

9 ISSUES HIGGS IN GLUON FUSION.5 LHC 8 TeV - ihixs.3 NNLO - PDF+α S uncertainties α s =.7,.9 α s=.7,.9 α s =.7, σ(h) [pb] NNPDF.3 MSTW8 CT DISCREPANCY NOT UNDERSTOOD DESPITE INTENSIVE BENCHMARKING CONSERVATIVE WAY OUT: TAKE THE ENVELOPE OF RESULTS (PDF4LHC PRESCRIPTION)

10 KNOWN ISSUES: METHODOLOGY EXAMPLE: THE d/u RATIO THE CMS W ASYMMETRY Ratio to NNPDF..4 THE d/u RATIO.3. d / u ( x, Q ) NNPDF. CT MSTW x LONG STANDING DISCREPANCY IN THE d/u RATIO BETWEEN MSTW AND OTHER GLOBAL FITS RESOLVED BY W ASYMMETRY DATA EXPLAINED BY INSUFFICIENTLY FLEXIBLE PDF PARAMETRIZATION NEW MSTW8DEUT SET

11 KNOWN ISSUES: METHODOLOGY PARAMETRIZATION MSTW8 DEUT VS DEFAULT CTX PDF SET (MSTW, 3) (P.Nadolsky, PDF4LHC, Nov.3, 4) IMPROVED PARAMETRIZATION (MORE PARAMETERS) MORE REALISTIC SHAPE MORE DATA BIGGER UNCERTAINTY

12 KNOWN ISSUES THEORY AND DATA TOP PRODUCTION 7 6 FFN, DIS VS DEFAULT LHC 8 TeV σ(tt) - top++v. NNLO+NNLL - α =.9 S NNPDF.3 ABM CT MSTW σ(tt) [pb] Global DIS FFN DIS+FFN Deut HT(p=) (CMS, 4) FFN SCHEME NOT ADEQUATE FOR HIGH ENERGY DATA DIS DATA RUNAWAY DIRECTION TOWARDS SMALL GLUON, SMALL α s CANNOT FIT TOP DATA

13 GETTING READY FOR RUN II: NNPDF3. THE NNPDF COLLABORATION: R. D. BALL, V. BERTONE, S. CARRAZZA, C. D. DEANS, L. DEL DEBBIO, A. GUFFANTI, M. P. HARTLAND, Z. KASSABOV, P. GROTH MERRILD, J. I. LATORRE, J. ROJO AND M. UBIALI

14 SUMMARY THE NNPDF METHODOLOGY ADDRESSES IN A QUANTITATIVE & SYSTEMATIC WAY ISSUES OF METHODOLOGY: CLOSURE TESTS THEORY CONSISTENCY: BENCHMARKS CASE STUDY I: JETS AND NNLO DATA: COMPATIBILITY & IMPACT ANALYSIS TOOLS CASE STUDY II: THE STRANGE PDF CASE STUDY III: CONSERVATIVE PARTONS CASE STUDY IV: ASSESSING DATA IMPACT

15 THE NNPDF METHODOLOGY A BRIEF SUMMARY MONTE CARLO APPROACH PSEUDODATA REPLICAS ARE GENERATED USING FULL INFO ON CORRELATED SYSTEMATICS MONTECARLO REPRESENTATION OF COVARIANCE MA TRIX A PDF REPLICA SET IS FITTED TO EACH PSEUDO DATA REPLICA THE SET OF PDF REPLICAS CAN BE USED TO COM PUTE ANY SET OF OBSERVABLES CENTRA VALUES (MEANS), UNCERTAINTIES (VARIANCES), CORRELATIONS (COVARIANCES), CONFIDENCE LEVELS... NEURAL NETWORK PARAMETRIZATION Experimental Data MC generation NN parametrization Fi i=,...,ndata NMC,BCDMS,SLAC,HERA,CHORUS... Fi() Fi() qnet() qnet() TRAINING Fi(N-) EVOLUTION Fi(N) qnet(n-) qnet(n) REPRESENTATION OF PROBABILITY DENSITY EACH PDF IS PARAMETRIZED BY A NEURAL NETWORK 37 PARAMETERS WITH DEFAULT 5 3 ARCHITECTURE MINIMIZATION IS PERFORMED USING GENETIC ALGORITHMS BEST FIT IS NOT MINIMUM OF χ (WOULD BE FITTING NOISE) CROSS VALIDATION DIVIDE DATA IN TWO SETS, FIT TRAINING SET, MONITOR VALIDATION SET, STOP WHEN VALIDATION χ STARTS RISING

16 THE NNPDF3. DATASET A BRIEF SUMMARY NEW IN NNPDF3. COMBINED HERA CHARM PRODUCTION (55 D.P) HERA II ZEUS+H STRUCTURE FUNCTIONS (778 D.P.) ATLAS.76TEV JETS (59 D.P.) ATLAS HIGH MASS DRELL YAN (5 D.P.) ATLAS W p T (9 D.P.) CMS W MUON ASYMMETRY ( D.P.) CMS DOUBLE DIFFERENTIAL DRELL YAN ( D.P.) CMS JETS (33 D.P.) CMS W +c ( D.P.) LHCB Z RAPIDITY (9 D.P.) ATLAS & CMS TOP TOTAL XSECT (3+3 D.P.) ] [ GeV / p / M Q 7 6 T FT DIS HERA FT DY TEV EW TEV JET ATLAS EW LHCB EW LHC JETS HERA ATLAS JETS.76TEV ATLAS HIGH MASS ATLAS WpT CMS W ASY CMS JETS CMS WC TOT CMS WC RAT LHCB Z TTBAR -5 NNPDF3. NLO dataset -4 x TOTAL DATASET: 476/478 (NLO/NNLO)

17 NNPDF3. THEORY AND METHODOLOGY A BRIEF SUMMARY NEW IN NNPDF3. UNCERTAINTIES: ALL CAN BE ADDITIVE OR MULTIPLICATIVE EW CORRECTIONS INCLUDED FULL MIGRATION OF CODE TO C++ PDF PARAMETRIZATION SCALE (Q = GeV) & BASIS (EVOLUTION EIGENSTATES) PREPROCESSING: BASIS INDEPENDENT, SELF CONSISTENTLY DETERMINED GENETIC ALGORITHM: NODAL MUTATION IMPLEMENTATION OF CROSS VALIDATION: LOOKBACK STOPPING

18 METHODOLOGY: CLOSURE TESTING

19 CLOSURE TESTS: THE BASIC IDEA ASSUME PDFS KNOWN: GENERATE FAKE EXPERIMENTAL DATA CAN DECIDE DATA UNCERTAINTY (ZERO, OR AS IN REAL DATA, OR... ) FIT PDFS TO FAKE DATA CHECK WHETHER FIT REPRODUCES UNDERLYING TRUTH : CHECK WHETHER TRUE VALUE GAUSSIANLY DISTRIBUTED ABOUT FIT CHECK WHETHER UNCERTAINTIES FAITHFUL CHECK STABILITY (INDEP. OF METHODOLOGICAL DETAILS)

20 ASSUME VANISHING EXPERIMENTAL UNCERTAINTY MUST BE ABLE TO GET χ = LEVEL CLOSURE TESTS UNCERTAINTY AT DATA POINTS TENDS TO ZERO (NOT NECESSARILY ON PDF!) DEFINE φ χ rep χ, EQUALS FIT UNCERTAINTY/DATA UNCERTAINTY; CHECK φ THE GLUON -5 BEST FIT ON TOP OF TRUTH IN DATA REGION x FRACTIONAL UNCERTAINTY VS TRAINING LENGTH χ VS TRAINING LENGTH Effectiveness of Genetic Algorithm in Level Closure Tests ) xg(x,q Level closure test vs. MSTW Level Closure Test Fit MSTW8nlo68cl Effectiveness of Genetic Algorithms in Level Closure Tests - Old (.3) genetic algorithm New genetic algorithm.45.4 NNPDF3. GA settings.35 NNPDF.3 GA settings - ] ].3 χ -3 fit ϕ [ T [f Number of Generations Number of GA Generations 4 5

21 LEVEL, LEVEL AND LEVEL LEVEL : FAKE DATA GENERATED WITH NO UNCERTAINTY INTERPOLATION AND EXTRAPOLATION UNCERTAINTY LEVEL : FAKE DATA GENERATED WITH SAME UNCERTAINTY AS REAL DATA (INCLUDING CORRELATIONS) LEVEL : NO PSEUDODATA REPLICAS: REPLICAS FITTED TO SAME DATA OVER AND OVER AGAIN FUNCTIONAL UNCERTAINTY DUE TO INFINITY OF EQUIVALENT MINIMA LEVEL : STANDARD NNPDF METHODOLOGY REPLICAS FITTED TO PSEUDODATA REPLICAS DATA UNCERTAINTY THREE SOURCES OF UNCERTAINTY COMPARABLE IN DATA REGION Ratios of gluon at different closure test levels THE GLUON: LEVEL, LEVEL AND LEVEL Lvl Closure Fit Lvl Closure Fit Lvl Closure Fit Ratios of gluon at different closure test levels Lvl Closure Fit Lvl Closure Fit Lvl Closure Fit x x

22 ) (x,q MSTW ) / g (x,q CT g LEVEL : CENTRAL VALUES AND UNCERTAINTIES CENTRAL VALUES: COMPARE FITTED VS. TRUE χ BOTH FOR INDIVIDUAL EXPERIMENTS & TOTAL DATASET FOR TOTAL χ =.±.3 UNCERTAINTIES: DISTRIBUTION OF DEVIA TIONS BETWEEN FITTED AND TRUE PDFS SAMPLED AT POINTS BETWEEN 5 AND FIND.699% FOR ONE SIGMA,.948% FOR TWO SIGMA C.L. THE GLUON: FITTED/ TRUE Ratio of Closure Test g to MSTW x LEVEL FITTED χ VS TRUE Distribution of χ for experiments NORM. DISTRIBUTION OF DEVIATIONS Distribution of single replica fits in level uncertainties.8 Closure test χ MSTW8nlo χ Closure test central χ MSTW8nlo central χ Replica distribution 8 Gaussian distribution χ D CDF DYE65 DYE886 HERAFCHARM HHERA ZEUSHERA HERAAV NTVDMN CHORUS BCDMS SLAC NMC Experiments TOP LHCB CMS ATLAS Entries Difference to theory (σ)

23 LEVEL STABILITY TESTS CHANGE UNDERLYING PDF SET (CT, NNPDF.3) INCREASE MAXIMUM GA TRAINING LENGTH TO 8K TESTS EFFICIENCY OF CROSS VALIDATION INCREASE NN ARCHITECTURE TO 5 NUMBER OF FREE PARAMETRES INCREASE BY MORE THAT CHANGE PDF PARAMETRIZATION BASIS OLD: ISOTRIPLET, ū d, s+ s, s s; NEW: ISOTRIPLET, SU(3) OCTET, BOTH TOTAL (q + q) & VALENCE (q q) STATISTICAL EQUIVALENCE! DISTANCES BETWEEN REF. AND NEW FIT: difference in unites of standard deviation of the mean 3K GENS VS 8K GENS.3 BASIS VS 3. BASIS 3 VS 37 PARMS

24 THEORY BENCHMARKING

25 CORRECTIONS AND CUTS PDFS PROVIDED AT LO, NLO, NNLO IN α s NNLO QCD CORRNS BENCHMARKED BETWEEN CODES: FOR DY MCFM VS FEWZ AT NLO; FEWZ & DYNNLO FOR NNLO; SIZE OF NNLO MONITORED: IN NLO FIT, DISCARD IF THAN EXPT UNCERTAINTIES LOWEST Q BIN OF CMS DDY DISCARDED dσ/dm(ll) per bin [pb] NLO, NNLO & EW CORRNS ATLAS AND CMS HIGH MASS DY ATLAS pp Z l + l - at LHC 7 TeV NNPDF3 NNLO NLO NNLO NLO+EW data dσ/dm(ll)/dy(ll) per bin [pb] CMS pp Z l + l - at LHC 7 TeV GeV < M ll < 3 GeV, NNPDF3 NNLO NLO NNLO NLO+EW data VIRTUAL PURE EW CORRECTIONS INCLUDED; FSR CORRECTED DATA USED; QED CORRECTIONS MONITORED HIGHEST Q BIN OF CMS DDY DISCARDED C-factors EW NNLO C-factors EW NNLO M(ll).5.5 Y(ll)

26 y JET DATA AT NLO AND NNLO PERCENTAGE ACCURACY OF THRESHOLD APP. CMS - δ for gg-channel IN GLUON CHANNEL y ATLAS 7 TeV - δ for gg-channel ) x g ( x, Q IMPACT OF THE JET DATA: NLO&NNLO NNPDF3. NLO, α S =.8, Q = GeV Global No Jet Data p (GeV) T p (GeV) T x y ATLAS.76 TeV - δ for gg-channel y CDF - δ for gg-channel ) x g ( x, Q NNPDF3. NNLO, α S =.8, Q = GeV Global No Jet Data p (GeV) T p (GeV) T x EXACT NNLO CORRECTIONS NOW KNOWN IN GLUON CHANNEL (GEHRMANN, GLOVER, DE RIDDER, PIRES, 4) THRESHOLD APPROX AVAILABLE IN ALL CHANNELS (DE FLORIAN ET AL, 4) ONLY ACCURATE AT HIGH p T, CENTRAL RAPIDITY, LARGER JET RADIUS (CARRAZZA & PIRES, 4) AT NNLO RETAIN ONLY BINS SUCH THAT ACCURACY BETTER THAN % CORRECTION THERE SMALLER THAN 5% ATLAS (R =.4) MOSTLY CUT OUT IMPACT OF JET DATA STILL SIZABLE

27 DATA CONSISTENCY & IMPACT ANALYSIS

28 NUCLEON STRANGENESS χ exp GLOBAL NO NEUTRINO NO W +c NO NEUTRINO/W +c CHORUS NUTEV ATLAS W, Z CMS W +c STRANGENESS RATIO AT LOW & HIGH SCALE Global Fit - NNLO, α S =.8, Q = GeV Global Fit no Neutrino Data no W+c Data no Neutrino, W+c Data 4 - NNLO, α S =.8, Q = GeV no Neutrino Data no W+c Data no Neutrino, W+c Data x IN GLOBAL FITS, STRANGENESS DETERMINED IN EQUAL PROPORTION BY NEUTRINO & W/Z PRODUCTION DATA NOW SPECIFIC INFORMATION FROM W +c BUT ONLY CMS DATAPOINTS; ATLAS NOT YET INCLUDED BECAUSE DATA ARE HADRON LEVEL DEFINE STRANGENESS RATIO r(x,q ) = s+ s ū+ d ALL DETERMINATIONS CONSISTENT WITHIN UNCERTAINTIES AT PRESENT NEUTRINO DATA CONTROL BOTH CENTRAL VALUE & UNCERTAINTY W/Z PRODUCTION DATA FAVOR LARGER r BUT WITH MUCH LARGER UNCERTAINTY CMS W +c DATA HAVE VERY LITTLE IMPACT, ON GLOBAL FIT OR ON FIT W/O NEUTRINO DATA (PERHAPS FAVOR HIGHER r) x

29 CONSERVATIVE PARTONS RESCALE ALL UNCERTAINTIES σ ασ: χ χ /α FOR A GIVEN EXPERIMENT DETERMINE PROBABILITY DISTRIBUTION P(α) (USING BAYES) DISCARD ALL EXPERIMENT FOR WHICH P(α) PEAKS WELL ABOVE ONE TWO OUT OF MEDIAN, MODE, MEAN, GREATER THAN α threshold χ =.9 FOR NNLO GOBAL, BECOMES χ =.6 FOR α threshold =.3, χ =. FOR α threshold =., χ =. FOR α threshold =., BUT CONSIDERABLE DETERIORATION OF UNCERTAINTIES ) x g ( x, Q ) x V ( x, Q CONSERV. VS. DEFAULT GLUON AND VALENCE -4 NNLO, α S =.8, Q = GeV -3 x - Global NNLO, α S =.8, Q = GeV Conservative, α max =. Conservative, α max =. - Global Conservative, α max =. Conservative, α max =. α PEAK FOR EXPERIMENTS DISCARDED IN CONS. FIT WHEN INCLUDED OR EXCLUDED FROM FIT NNLO global fit NNLO cons. fit αmax =. Experiment mean mode median mean mode median NMC σ NC,p SLAC BCDMS CHORUS ZEUS HERA-II H HERA-II HERA σ NC c E886 p CDF Z rapidity CDF Run-II k t jets ATLAS W, Z ATLAS high-mass DY CMS W muon asy CMS W +c total CMS W +c ratio CMS D DY LHCb x - -

30 ) [ref] ) [new] / g ( x, Q g ( x, Q ) [ref] ) [new] / d ( x, Q d ( x, Q THE IMPACT OF LHC (AND HERA) DATA OVERALL MEASURE OF IMPACT: φ FIT UNCERTAINTY/DATA UNCERTAINTY HERA II IMPACT SIZABLE IMPACT OF LHC DATA MODERATE BUT VISIBLE IMPACT OF CMS OR ATLAS COMPARABLE TO (MOD ERATE) IMPACT OF NON LHC, NON HERA DATA GLOBAL VS NO LHC: g& d -4-4 Global Fit -3 x - 4 NNLO, α S =.8, Q = GeV No LHC data Global Fit -3 x NNLO, α S =.8, Q = GeV No LHC data - ) [ref] ) [new] / g ( x, Q g ( x, Q ) [ref] ) [new] / d ( x, Q d ( x, Q GLOBAL VS HERA+ATLAS: g& d x 4 NNLO, α S =.8, Q = GeV HERA-only HERA + ATLAS Global fit -3 - x 4 - NNLO, α S =.8, Q = GeV HERA-only HERA + ATLAS Global fit - FRACTIONAL UNCERTAINTY Dataset ϕ χ NLO ϕ χ NNLO Global.9.3 HERA-I HERA all HERA+ATLAS HERA+CMS Conservative no LHC.3.36 ) [ref] ) [new] / g ( x, Q g ( x, Q ) [ref] ) [new] / d ( x, Q d ( x, Q GLOBAL VS HERA+CMS: g& d x 4 NNLO, α S =.8, Q = GeV HERA-only HERA + CMS Global fit -3 - x 4 - NNLO, α S =.8, Q = GeV HERA-only HERA + CMS Global fit -

31 IMPACT OF THE COMBINED HERA DATA REPLACE ALL HERA DATA WITH NEW COMBINED SET φ OF HERA NON COMBINEDCDATA QUITE SMALL SMALL IMPACT φ OF COMBINED HERA & TOTAL COMBINED HERA QUITE SIMILAR IMPACT OF COMBINATION SMALL

32 DATA: WHERE LIES THE PROMISE CAN USE α & φ INDICATORS TO ASSESS OBJECTIVELY WHICH DATA COULD HAVE BIGGER IMPACT LARGE α MODE MUST IMPROVE COMPATIBILITY SMALL φ MUST REDUCE UNCERTAINTY Value P(α) MODE & φ Distribution of φ and mode P(α) for experiments NNPDF3. NLO φ Mode P(α) NNLO global fit Experiment N dat P(α) mode φ ATLAS W, Z 3..3 ATLAS 7 TeV jets ATLAS.76 TeV jets ATLAS high-mass DY CMS W electron asy.87.9 CMS W muon asy.4.5 CMS jets CMS W +c total CMS W +c ratio CMS D DY LHCb W, Z rapidity 9..4 σ(t t) NMC HHERA ZEUSHERA HERAAV NTVDMN CHORUS BCDMS SLAC CDF DYE65 DYE886 HERAFCHARM D TOP LHCB CMS ATLAS Experiments SOME PRELIMINARY INDICATIONS....76TEV JET DATA SEEEM TO HAVE OVERESTIMATED UNCERTAINTIES, POTENTIALLY HIGH IMPACT TOP DATA: SUPRISINGLY HIGH IMPACT GIVEN SMALL DATA SAMPLE HIGH MASS DY CONSISTENCY PROBLEM

33 OUTLOOK WHAT LIES BEHIND THE CORNER

34 SUMMARY ALL GLOBAL FITTING GROUPS HAVE PUBLISHED MAJOR UPDATES FOR LHC RUN II: METHODOLOGY AND DATA IMPROVEMENT IMPROVED AGREEMENT TOWARDS A NEW PDF4LHC PRESCRIPTION: COMBINED PDF SETS NEW ANALYSIS TOOLS: CMC, META/MCH AND SM PDFS

35 GLOBAL FITS: PROGRESS MMHT (SUCCESSOR OF MSTW8) DECEMBER 4 CT4 JUNE 5 FOR ALL, PROGRESS IN METHODOLOGY & DATASET METHODOLOGY NNPDF3. MMHT4 CT4 NO. OF FITTED PDFS PARAMETRIZATION NEURAL NETS x a ( x) b CHEBYSCHEV x a ( x) b BERNSTEIN FREE PARAMETERS UNCERTAINTIES REPLICAS HESSIAN HESSIAN TOLERANCE NONE DYNAMICAL DYNAMICAL CLOSURE TEST REWEIGHTING REPLICAS EIGENVECTORS EIGENVECTORS MMHT, CT LARGER # OF PARMS., ORTHOGONAL POLYNOMIALS NNPDF CLOSURE TEST

36 DATASET NNPDF3. MMHT4 CT4(PREL) SLAC P,D DIS BCDMS P,D DIS NMC P,D DIS E665 P,D DIS CDHSW NU DIS CCFR NU DIS CHORUS NU DIS CCFR DIMUON NUTEV DIMUON HERA I NC,CC HERA I CHARM H,ZEUS JETS H HERA II ZEUS HERA II E65 & E866 FT DY CDF & D W ASYM CDF & D Z RAP CDF RUN II JETS D RUN II JETS ATLAS HIGH MASS DY CMS D DY ATLAS W,Z RAP ATLAS W PT CMS W ASY CMS W +C LHCB W,Z RAP ATLAS JETS CMS JETS TTBAR TOT XSEC TOTAL NLO TOTAL NNLO

37 Quark - Antiquark Luminosity Gluon - Gluon Luminosity PROGRESS... PARTON LUMINOSITIES: IMPROVED AGREEMENT QUARK QUARK LHC 8 TeV - Ratio to NNPDF.3 NNLO - NNPDF.3 NNLO CT NNLO MSTW8 NNLO M X LHC 8 TeV - Ratio to NNPDF.3 NNLO - NNPDF.3 NNLO CT NNLO MSTW8 NNLO M X α s 3 α s 3 =.8 =.8 Ratio GLUON GLUON Ratio Quark-Quark, luminosity nnpdf3.nnlo mmhtnnlo ct4nnlo S = 8.e+3 GeV M X M X [GeV] [GeV] 3 Gluon-Gluon, luminosity nnpdf3.nnlo mmhtnnlo ct4nnlo S = 8.e+3 GeV 3 Generated with APFEL.4. Web Generated with APFEL.4. Web

38 PROGRESS: STANDARD CANDLES & HIGGS PDF UNCERTAINTY ON HIGGS PRODUCTION DOWN TO ABOUT % CONSERVATIVE ENVELOPE NO LONGER NEEDED STATISTICAL COMBINATION ADEQUATE HIGGS IN GLUON FUSION ggh, gghiggs NNLO, LHC 3 TeV, α s =.8 - W + PRODUCTION W, VRAP NNLO, LHC 3 TeV, α s =.8 Cross-Section (pb) NNPDF3. MMHT4 CT4p CMCPDF Envelope Cross-Section (nb) NNPDF3. MMHT4 CT4p CMCPDF Envelope (Carrazza, Latorre, Rojo, Watt, 5)

39 TOWARDS A NEW PDF4LHC PRESCRIPTION PERFORM MONTE CARLO COMBINATION OF UNDERLYING PDF SETS SETS ENTERING THE COMBINATION MUST SATISFY COMMON REQUIREMENTS DELIVER A SINGLE COMBINED PDF SET THROUGH SUITABLE TOOLS

40 MONTE CARLO COMBINATION CONVERT ALL SETS INTO MONTE CARLO (Watt, S.F., -3) HESSIAN SETS CAN BE CONVERTED BY PERFORMING MONTE CARLO IN PARAMETER SPACE (Watt, Thorne, ) COMBINE MONTE CARLO REPLICAS INTO SINGLE SET COMBINED MC SETS FOR ANTIDOWN & STRANGE (Carrazza, Latorre, Rojo, Watt, 5)

41 CMC PDFS AND NONGAUSSIAN BEHAVIOUR DEVIATION FROM GAUSSIANITY OBSERVED E.G. AT LARGE x, DUE TO LARGE UNCERTAINTY & POSITIVITY BOUNDS (MAY BE RELEVANT FOR SEARCHES) MONTE CARLO COMPARED TO HESSIAN CMS W + c production LHCb electrons MONTE CARLO PDFS REPRODUCE NONGAUSSIAN, HESSIAN FAIL PROBLEM DOES NOT ARISE FOR BULK OF DATA HESSIAN ADEQUATE!

42 TOOLS FOR DELIVERY COMBINED MC SET IS LARGE (3 REPLICAS) HESSIAN DELIVERY OFTEN DESIRABLE (PDF UNCERTAINTIES AS NUISANCE PARAMETERS) CMC: GA COMPRESSION OF MC SET TO A SMALLER SET WITH MINIMAL INFORMATION LOSS (3 4 REPLICAS) META PDFS OR MC H PDFS: HESSIAN REPRESENTATION OBTAINED BY REFITTING MC REPLICAS WITH FUNCTIONAL FORM (META), OR REPRESENTING THEM ON LINEAR BASIS OF REPLICAS (MC H) VALIDATION AND BENCHMARKING SUCCESFUL GLUON GLUON CORRELATION STARTING COMBINED SET COMPRESSED MC SET META PDF SET (PDF4LHC: preliminary, 5)

43 THE MCH IDEA: NONGAUSSIAN EFFECTS NOT VERY IMPORTANT FOR BULK OF DATA HESSIAN DELIVERY OFTEN ADVANTAGEOUS FOR ANALYSIS CAN WE CONVERT MC TO HESSIAN (GAUSSIAN)? A: USE REPLICAS AS BASIS FUNCTIONS (CONTINUITY, SUM RULES, DGLAP AUTOMATICALLY IMPLEMENTED) MCH GA SINGLE OUT GAUSSIAN REGION (ONE SIGMA = 68% C.L.) OPTIMIZE UNCERTAINTIES VIA GENETIC ALGORITHM MCH PCA (Carrazza, S.F., Rojo, Kassabov 5) REPRODUCE COVARIANCE MATRIX ON VERY FINE SET OF GRID POINTS REPRESENT IT BY SINGULAR VALUE DECOMPOSITION ON REPLICAS PICK LARGEST CONTRIBUTIONS (PRINCIPAL COMPONENT ANALYSIS) COVARIANCE MATRIX PERFECTLY REPRODUCED; NO CHOICES: NO BIAS

44 MCH PCA PRIOR COMBINATION OF CT4, MMHT, NNPDF3.: MC9 OR META3 MCH=PCA: SETS OF OR HESSIAN EIGENVECTORS META PDFS: V SETS OF *3 AND *5 EIGENVECTORS THE CORRELATION MATRIX: PRIOR FINAL (HIGH RESOLUTION) MCH MCH ACHIEVES PERCENT ACCURACY

45 MINIMAL SETS: SM PDFS GIVEN ONE OR MORE PROCESSS ( SIGNAL ), SELECT EIGENVECTORS WHICH PROVIDE THE DOMINANT CONTRIBUTION: DETERMINE SUBSET OF GRID POINTS WHICH HAVE THE HIGHEST CORRELATION TO THE PROCESS PERFORM PCA ANALYSIS ON THE CORRESPONDING SUBMATRIX NOTE USAGE OF SUBGRID GUARANTEES STABILITY: SIMILAR PROCESSES WILL ALSO BE WELL REPRODUCED (EXAMPLE: TOP VS HIGGS).. HIGGS 3 eigenvectors HIGGS p T DISTRIBUTION IN GLUON FUSION

46 MINIMAL SETS: SM PDFS GIVEN ONE OR MORE PROCESSS ( SIGNAL ), SELECT EIGENVECTORS WHICH PROVIDE THE DOMINANT CONTRIBUTION: DETERMINE SUBSET OF GRID POINTS WHICH HAVE THE HIGHEST CORRELATION TO THE PROCESS PERFORM PCA ANALYSIS ON THE CORRESPONDING SUBMATRIX NOTE USAGE OF SUBGRID GUARANTEES STABILITY: SIMILAR PROCESSES WILL ALSO BE WELL REPRODUCED (EXAMPLE: TOP VS HIGGS) ADD ONE OR MORE PROCESSES ( BACKGROUND ): REPEAT ANALYSIS & PICK FURTHER EIGENVECTORS. HIGGS+TOP 4 eigenvectors TOP INVARIANT MASS DISTRIBUTION

47 MINIMAL SETS: SM PDFS GIVEN ONE OR MORE PROCESSS ( SIGNAL ), SELECT EIGENVECTORS WHICH PROVIDE THE DOMINANT CONTRIBUTION: DETERMINE SUBSET OF GRID POINTS WHICH HAVE THE HIGHEST CORRELATION TO THE PROCESS PERFORM PCA ANALYSIS ON THE CORRESPONDING SUBMATRIX NOTE USAGE OF SUBGRID GUARANTEES STABILITY: SIMILAR PROCESSES WILL ALSO BE WELL REPRODUCED (EXAMPLE: TOP VS HIGGS) ADD ONE OR MORE PROCESSES ( BACKGROUND ): REPEAT ANALYSIS & PICK FURTHER EIGENVECTORS CAN ITERATE AT WILL EVENTUALLY RECOVER FULL EIGENVECTOR SET HIGGS+TOP+Z+W 5 eigenvectors W p T DISTRIBUTION

48 OUTLOOK: NNPDF3. NEW DATA (MINIMAL SET) HERA II COMBINED DATA arxiv:56.64 D W ASYMMETRY arxiv:4.86 ATLAS LOW MASS DY arxiv:44. ATLAS PROMPT PHOTON arxiv:3.44 ATLAS W +c arxiv:4.663 ATLAS Z p t arxiv: ATLAS t t RAPIDITY arxiv:47.37 ATLAS INCLUSIVE JETS 7 TEV 5 FB arxiv: CMS DOUBLE DIFFERENTIAL 8 TEV arxiv:4.5 CMS Z p t arxiv:54.35 LHCB W µν RAP. arxiv:

49 NNPDF3. AND BEYOND RESUMMED PDFS SMALL x AND EVOLUTION THRESHOLD RESUMMATION AND NEW PHYSICS SEARCHES HEAVY QUARKS RUNNING QUARK MASSES & QUARK MASS STUDIES FITTED ( INTRISIC ) CHARM THEORY UNCERTAINTY ON PDFS SCALE VARIATION PERTURBATIVE SERIES: CACCIARI HOUDEAU MORE CLOSURE TESTS: INCONSISTENT DATA QUANTITATIVE MODELS OF FUNCTIONAL UNCERTAINTY TOLERANCE & χ RULE QED PDF SET MIXED QCD QED EFFECTS DETERMINATION OF PHOTON FROM WW PRODUCTION MONTE CARLO PDFS

50 PDFS: PLUMBING OR PRECISION PHYSICS?

51 EXTRAS

52 TREATMENT OF α s NEW SIMPLIFIED PDF4LHC PRESCRIPTION SEPARATE α s UNCERTAINTY FROM PDF UNCERTAINTY: DETERMINE EACH SEPARATELY AND COMBINE IN QUADRATURE IF NEEDED PROVEN TO BE EQUIVALENT TO CORRELATED DETERMINATION, UP TO NONLINEAR TERMS AGREE ON A CENTRAL VALUE AND AN UNCERTAINTY ON α s (SEE BELOW) FOR THE DETERMINATION OF THE PDF UNCERTAINTY, USE PDFS AT THE (FIXED) CENTRAL VALUE OF α s FOR THE DETERMINATION OF THE α s UNCERTAINTY, COMPUTE OBSERVABLES WITH α s SHIFTED BY ONE σ, WITH PDFS CORRESPONDING TO THE GIVEN FIXED VALUE TAKE RESULT AS ONE σ α s UNCERTAINTY ON OBSERVABLE

53 THE VALUE OF α s PDG VALUE (AUGUST 4): α s (M Z ) =.85±.6 COMMENTS LATTICE UNCERTAINTY CURRENTLY ESTIMATED BY FLAG (arxiv:3.8555) TO BE TWICE THE PDG VALUE (±.) IT IS AN AN AVERAGE OF AVERAGES SOME SUB AVERAGES (E.G. DIS) INCLUDE MU TUALLY INCONSISTENT/INCOMPATIBLE DATA/EXTRACTIONS SOME SUB AVERAGES (E.G. τ OR JETS) INCLUDE DETERMINATIONS WHICH DIFFER FROM EACH OTHER BY EVEN FOUR FIVE σ AVERAGING THE TWO MOST RELIABLE VALUES (GLOBAL EW FIT & τ, BOTH N 3 LO, NO DEP. ON HADRON STRUCTURE) GIVES α s =.96±. NEW PDF4LHC AGREEMENT PDG UNCERTAINTY CONSERVATIVELY MULTIPLIED BY CENTRAL VALUE & UNCERTAINTY ROUNDED: PDF SETS USUALLY GIVEN IN STEPS OF α s (M z ) =. α s (M Z ) =.8±.

54 THE NNPDF APPROACH: THE NEURAL MONTE CARLO BASIC IDEA: MONTE CARLO SAMPLING OF THE PROBABILITY MEASURE IN THE (FUNCTION) SPACE OF PDFS START FROM MONTE CARLO SAMPLING OF DATA SPACE SPACE OF FUNCTIONS HUGE 5 BINS FOR PTS 7 FCTNS BINS IMPORTANCE SAMPLING: DATA TELL US WHICH BINS ARE POPULATED replica averages vs. central values Monte Carlo replicas.. NNPDF. - Central values N rep = N rep = N rep =.... Experimental data replica standard dev. vs. uncertainties Monte Carlo replicas... N rep = N rep = N rep = NNPDF. - Errors..... Experimental data REPLICAS ENOUGH FOR CENTRAL VALS, FOR UNCERTAINTIES, FOR CORRELNS Experimental Data MC generation NN parametrization Fi i=,...,ndata NMC,BCDMS,SLAC,HERA,CHORUS... Fi() Fi() qnet() qnet() TRAINING Fi(N-) EVOLUTION Fi(N) qnet(n-) qnet(n) REPRESENTATION OF PROBABILITY DENSITY

55 DATA MONTE CARLO PDF MONTE CARLO NEURAL NETWORK PARM+ CROSS VALIDATION METHOD EACH PDF NEURAL NETWORK PARAMETRIZED BY 37 PARAME TERS NNPDF.: 37 7 = 59 PARMS (RECALL MSTW, CTEQ FREE PARAMETERS) INFINITE NUMBER OF PARAM ETERS CAN REPRESENT ANY FUNCTION COMPLEX SHAPES (LARGE NO.OF PARAMETERS) REQUIRE LONGER FITTING FIT STOPS WHEN QUALITY OF FIT TO RANDOMLY SELECTED VALIDATION DATA (NOT FITTED) STOPS IMPROVING CAN OBTAIN A FIT WITH χ LOWER THAN BEST FIT ( OVERLEARNING ) Experimental Data MC generation NN parametrization Fi i=,...,ndata NMC,BCDMS,SLAC,HERA,CHORUS... Fi() Fi() qnet() qnet() TRAINING Fi(N-) EVOLUTION Fi(N) qnet(n-) qnet(n) REPRESENTATION OF PROBABILITY DENSITY

56 DATA MONTE CARLO PDF MONTE CARLO CROSS VALIDATION REPLICAS ARE FITTED TO A DATA SUBSET A DIFFERENT SUBSET OF DATA USE FOR EACH REPLICA OPTIMAL FIT WHEN FIT TO VALIDATION (CONTROL) DATA STOPS IMPROVING χ OPTIMAL FITTING FIT TO DATA

57 DATA MONTE CARLO PDF MONTE CARLO CROSS VALIDATION REPLICAS ARE FITTED TO A DATA SUBSET A DIFFERENT SUBSET OF DATA USE FOR EACH REPLICA OPTIMAL FIT WHEN FIT TO VALIDATION (CONTROL) DATA STOPS IMPROVING THE BEST FIT IS NOT AT THE MINIMUM OF THE χ χ OVERFITTING FIT TO DATA

58 LOOKBACK CROSS VALIDATION IN PREVIOUS NNPDF RELEASES, INCREASING AND DECREASING TRAINING AND VALIDATION χ DEFINED IN TERMS OF THRESHOLD VALUES δ tr AND δ val : INCREASE: χ val (N gen + ) > χ (N gen + )+δ val IN NNPDF3. USE LOOKBACK: FIT IS RUN UP TO SOME LARGE # OF GA GENERATIONS THEN ONE LOOKS BACK FOR ABSOLUTE MIN. OF VALIDATION χ CHECK THAT RESULTS ARE INDEPENDENT OF THE LARGE # OF GA GENS CHECK THAT RESULTS ARE INDEPENDENT OF FLUCTUATIONS IN VALUE OF ABSOLUTE MINIMUM DIFFERENT STOPPING POINTS, BUT INDISTINGUISHABLE PDFS

59 Entries Replica distribution Distribution of single replica fits in level uncertainties Difference to theory (σ) LEVEL VS. LEVEL DO WE NEED THE PSEUDODATA REPLICAS? TRUTH FIT: LEVEL VS. LEVEL xg(x,q ) Gaussian distribution Level Reweighted Level Fitted Entries Replica distribution Distribution of single replica fits in level uncertainties Gaussian distribution Difference to theory (σ) REWEIGHT VS REFIT: LEVEL VS. LEVEL xg(x,q ) Level Reweighted Level Fitted LEVEL UNDERESTIMATES UNCERTAINTIES LEVEL TRUTH FIT DISTN. WIDER THAN PREDICTED REWEIGHT VS. REFIT: X STAT. EQUIVALENCE FOR LEVEL REWEIGHT VS. REFIT: REFITTED SIGMA TOO SMALL FOR LEVEL REWEIGHT REFIT Entries (LEVEL ) Distribution of distance for Central Values.6 LVL reweighted vs refit χ Distribution for DOF x x x

60 uv alpha effective exponent -4 x ASYMPTOTIC EXPONENTS -3 ClosureTest_MSTW_lvl_k_fl, 68% c.l. MSTW8nlo68cl, central value - LEVEL TEST uv beta effective exponent ClosureTest_MSTW_lvl_k_fl, 68% c.l. MSTW8nlo68cl, central value x AS x, PDF GOES AS x α, AS x, PDF GOES AS ( x) β DEFINE EFFECTIVE EXPONENTS α eff,i (x) lnf i(x) ln/x, β eff,i (x) lnf i(x) ln( x) COMPARE VALUES FROM FIT WITH VALUES OBTAINED FROM UNDERLYING TRUE (MSTW8) FORM

61 Large-x Effective Exponent for Gluon PREPROCESSING Small-x effective exponent for the Singlet x x -3 - PDFS ARE PARAMETRIZED WITH NEURAL NETWORKS TIMES PREPROCESSING FUNCTION: f i (x) = x α i ( x) β i NN(x) GOAL IS TO SPEED UP TRAINING WITHOUT BIASING RESULT α i, β i RANDOM REPLICA BY REPLICA WITH UNIFORM DISTRIBUTION IN RANGE CHECK THAT THE RANGE IS WIDER THAN THE 9% C.L. OF EFFECTIVE EXPONENTS

62 Quark - Antiquark Luminosity NNPDF3. VS CT & MSTW8 LHC 3 TeV, NNLO, α S (M )=.8 - Ratio to NNPDF.3 NNPDF3. CT MSTW8 Z PARTON LUMINOSITIES Quark - Quark Luminosity LHC 3 TeV, NNLO, α S (M )=.8 - Ratio to NNPDF.3 NNPDF3. CT MSTW8 Z.85 M X ( GeV ) 3.85 M X ( GeV ) 3 Quark - Gluon Luminosity LHC 3 TeV, NNLO, α S (M )=.8 - Ratio to NNPDF.3 NNPDF3. CT MSTW8 Z Gluon - Gluon Luminosity LHC 3 TeV, NNLO, α S (M )=.8 - Ratio to NNPDF.3 NNPDF3. CT MSTW8 Z.85 M X ( GeV ) 3.85 M X ( GeV ) 3

63 Quark - Antiquark Luminosity LHC 3 TeV, NNLO, α S (M )=.8 - Ratio to NNPDF.3 NNPDF.3 NNPDF3. Z NNPDF.3 VS NNPDF3. PARTON LUMINOSITIES Quark - Quark Luminosity LHC 3 TeV, NNLO, α S (M )=.8 - Ratio to NNPDF.3 NNPDF.3 NNPDF3. Z.85 M X ( GeV ) 3.85 M X ( GeV ) 3 Quark - Gluon Luminosity LHC 3 TeV, NNLO, α S (M )=.8 - Ratio to NNPDF.3 NNPDF.3 NNPDF3. Z Gluon - Gluon Luminosity LHC 3 TeV, NNLO, α S (M )=.8 - Ratio to NNPDF.3 NNPDF.3 NNPDF3. Z.85 M X ( GeV ) 3.85 M X ( GeV ) 3

64 ) [ref] ) [new] / g ( x, Q g ( x, Q NNLO, α S =.8, Q NNPDF.3 NNLO NNPDF3. NNLO NNPDF.3 VS NNPDF3. METHODOLOGY VS. DATA NNLO.3 VS 3.: GLUON & VALENCE 4 = GeV ) [ref] ) [new] / V ( x, Q V ( x, Q NNLO, α S =.8, Q NNPDF.3 NNLO NNPDF3. NNLO 4 = GeV x x DEFAULT VS.3 LIKE DATASET: GLUON & ANTIDOWN NNLO, α S =.8, Q 4 = GeV NNLO, α S =.8, Q 4 = GeV ) [ref] ) [new] / g ( x, Q. Global Fit.5.3 Dataset ) [ref]..5 ) [new] / d ( x, Q g ( x, Q Global Fit.5.3 Dataset..5 d ( x, Q x x - -

65 ) x g ( x, Q NNPDF3., α S =.8, Q PERTURBATIVE STABILITY LO, NLO, NNLO: GLUON & VALENCE = GeV LO NLO NNLO ) x V ( x, Q.5.5 NNPDF3., α S =.8, Q = GeV LO NLO NNLO ) ) ) / g ( x, Q ( g ( x, Q x x - - NNPDF3., α S =.8, Q = GeV NLO -sigma PDF uncertainty NNLO -sigma PDF uncertainty ( NNLO - NLO ) / NNLO NLO, NNLO PDF UNC. & NLO TH. UNC ) ) ) / V ( x, Q ( V ( x, Q x x NNPDF3., α S =.8, Q = GeV NLO -sigma PDF uncertainty NNLO -sigma PDF uncertainty ( NNLO - NLO ) / NNLO -

66 MULTIPLICATIVE VS. ADDITIVE UNCERTAINTIES NNPDF3. NNLO, α s =.8, Q = GeV NNPDF3. NNLO, α s =.8, Q = GeV ).5 Baseline Additive ).4. Baseline Additive x g ( x, Q x Σ ( x, Q x x NORMALIZATION UNCERTAINTY PROPORTIONAL TO MEASURED VALUE MULTIPLICATIVE WHAT ABOUT OTHER UNCERTAINTIES? IN COLLIDER MOSTLY MULTIPLICATIVE (A. Glazov) IN NNPDF3. DEFAULT ALL UNCERTAINTIES MULTPLICATIVE FOR HERA, TEVATRON, LHC REPEAT FIT WITH ALL UNCERTAINTIES ADDITIVE (AS IN NNPDF.3) t METHOD USED FOR MULTIPLICATIVE UNCERTAINTIES NEGLIGIBLE CHANGE SEEN

67 REQUIREMENTS FOR INCLUSION PDFS TO BE INCLUDED IN THE COMBINATION MUST SATISFY REQUIREMENTS WHICH MAKE THEM COMPATIBLE DOCUMENT IN PDF4LHC APRIL 5 INDICO SOME LISTED HERE (SEE DOCUMENT FOR FULL LIST): PROVIDE RESULTS AT NLO AND NNLO, WITH NNLO α s, EVOLUTION BENCHMARKED AGAINST HOPPET OR QCDNUM SEPARATE TREATMENT OF α s AS DISCUSSED USE A GENERAL MASS VARIABLE FLAVOR NUMBER SCHEME WITH UP TO 5 FLAVORS CURRENT TECHNOLOGY: FITS MUST BE BASED ON APPROXIMATELY COMMON, GLOBAL SET OF DATA REQUEST TO RELAX THIS (HERAPDF)