Update of comparison of QCD fits 29/01/2008 Joël Feltesse 1 Outlook • Double minimum with H1 param • Double (triple) minimum with ZEUS-Jet and Inbetween parametrisations • Last comparisons between results of JF (H1fitter) and ACS programmes. Joël Feltesse 2 H1-ZEUS data set. Gluon at Q2 = 4 GeV2 for various input param. Ref Inbetween H1-Par. data set H1-Z With error band Zeus-S Zeus-J Large difference with H1-par. Gluon!! 20/11/2007 Joël Feltesse 3 Reminder • JF :Two types of solutions pending on the parameteristion: – Humpy gluon with H1param and Inbetween – Straight gluon at low x with ZEUS-J and ZEUS-S • Mandy : one only type of solution for all 4 parameterisations: – Straight(smooth) gluon at low x Joël Feltesse 4 New test • Use H1 parametrisation but giving as input to Minuit the initial values of the parameter of a straight gluon. • Surprising results – Initial straight gluon gives after minimisation a straight gluon : Chi2 = 446 [fit_2] – Initial humpy gluon gives after mininisation a humpy gluon : Chi2 = 456 [[fit_1] – No way (so far) to jump by minimisation from the humpy gluon to to the other !! – Double minimum also observed with QCDFIT (Li) and with ZEUS package (Mandy) Joël Feltesse 5 Joël Feltesse 6 Joël Feltesse 7 Joël Feltesse 8 Comparison of output values of parameters at Q02 = 4 GeV2 JF output, H1 Param. H1PDF2k init. values Fit_1, chi2 = 456.4 NO. NAME VALUE ERROR 1 Bg -0.93489 0.38909E-01 2 Cg 9.2298 0.73734 3 Dg 11398. 8871.4 4 BU -0.21035 0.79069E-02 5 CU 4.8602 0.17934 6 FU 274.36 52.835 7 AD 0.15956 0.88072E-02 8 CD 4.0076 0.33801 9 CUbar 6.2007 0.63037 10 CDbar 5.8054 1.3173 JF output, H1-Param with mandy init.values Fit_2, chi2 = 446.0 NAME VALUE ERROR 1 Bg -0.85283E-01 0.35957E-01 2 Cg 0.13562E+02 0.12354E+01 3 Dg 0.16840E+02 0.65782E+01 4 BU -0.20143E+00 0.40238E-02 5 CU 0.48627E+01 0.19240E+00 6 FU 0.26490E+03 0.54386E+02 7 AD 0.17032E+00 0.53413E-02 8 CD 0.40289E+01 0.32759E+00 9 CUbar 0.72805E+01 0.50364E+00 10 CDbar 0.47830E+01 0.11619E+01 Where PDF parametrisation : x f(x) = A xB (1 - x)C (1 + D x + F x3 ) Joël Feltesse 9 Obvious questions • Is the double minimum feature unique to H1 parameterisation ? • Is the double minimum unique to H1/ZEUS combined data set ? • Is the chi2 difference always in favor of a straight gluon ? Joël Feltesse 10 Double minimum with ZEUS-Jet parameterisation (and ZEUS HQ treatment) ? – Initial straight gluon gives after minimisation a straight gluon : Chi2 = 443.9 – Initial humpy gluon gives after minimisation a humpy gluon : Chi2 = 456.3 → Results similar to H1 parametrisation Joël Feltesse 11 Joël Feltesse 12 Double minimum with Inbetween parameterisation (and H1 HQ treatment) ? – Initial straight gluon gives after minimisation a straight gluon : Chi2 = 442.3 – Initial humpy gluon gives after mininisation a humpy gluon : Chi2 = 433.8 ! the smallest Chi2 so far BUT → – Dvalence negative at large x is an unphysical solution (Mandy). It even gives a negative CC x-sec at large x ! – Initial humpy gluon and dvalence density forced to be positive gives after minimisation a new minimum with a humpy gluon : Chi2 = 450.1 ! Joël Feltesse 13 Joël Feltesse 14 Joël Feltesse 15 Double minimum with only H1 data (from H1PDF2k) and H1 parameterisation ? – Initial straight gluon gives after minimisation a straight gluon : Chi2 = 547.3 – Initial humpy gluon gives after mininisation a humpy gluon : Chi2 = 536.8. Smaller than the straight gluon minimum (relief). – Remark : as for the publication the fit has been performed with full correlation taken into account. Joël Feltesse 16 In short When fitting H1Z combined data sets Minuit minimisation finds two minimum independently of the parameterisation (H1-param., Inbetween, ZEUS-Jet). Each time, the straight gluon minimum is slightly favored (Chi2 smaller by 6 to 13 units) provided unphysical solutions are removed. Is the double minimum a problem ? Is there a third minimum ? I feel uneasy than Minuit is not capable to jump alone from a minimum to a better one. The double minimum is also observed when using H1 published data sets but then favoring the humpy solution. → Better find a way to get one only robust minimum (Li’s talk) Joël Feltesse 17 Considering only the straight gluon solutions. How different are the 3 fits ? Reminder. All fits with : Q2min = 3.5 GeV2, 573 data points and all errors uncorrelated. H1 param (10 parameters), chi2 = 446.1 Inbetween (12 parameters), chi2 = 442.3 ZEUS-JET (11 parameters), chi2 = 443.9 Chi2 are very close, but PDFs are not so close → Joël Feltesse 18 Joël Feltesse 19 Joël Feltesse 20 Joël Feltesse 21 Staight Gluon. Technical comparison between JF and Mandy ZEUS J Parameterisation Chi2_J = 443.9 Chi2_M = 440.8 Val_J Err_J Val_M Err_M Dif/Err Gluon Bg -0.092 Gluon Cg 12.934 Gluon Dg 15.036 u valence Buv 0.590 u valence Cuv 3.817 u valence Duv 2.450 d valence Cdv 4.866 d valence Ddv 2.597 Sea Asea 0.567 Sea Bsea -0.210 Sea Csea 3.662 0.028 -0.104 0.844 13.323 4.350 17.633 0.037 0.616 0.121 3.825 0.755 2.103 0.726 4.803 2.030 2.089 0.019 0.582 0.004 -0.206 0.541 3.743 Joël Feltesse 0.034 0.925 5.879 0.039 0.138 0.744 0.910 2.362 0.022 0.004 0.655 0.396 -0.440 -0.508 -0.693 -0.067 0.463 0.077 0.677 -0.735 -0.712 -0.135 22 Staight Gluon. Technical comparison between JF and Mandy H1 Parameterisation Chi2_J = 446. Val_J Gluon Bg Gluon Cg Gluon Dg B_U C_U F_U A_D C_D C_Ubar C_Dbar Err_J -0.085 0.036 13.562 1.235 16.840 6.578 -0.201 0.004 4.863 0.192 264.900 54.386 0.170 0.005 4.029 0.328 7.280 0.504 4.783 1.162 Joël Feltesse Chi2_M = 439.3 Val_M Err_M Dif/Err -0.090 13.372 16.648 -0.201 4.882 268.61 0.172 4.016 7.724 4.584 0.035 0.290 0.425 0.004 0.096 6.951 0.005 0.323 0.520 0.157 0.139 0.154 0.029 -0.218 -0.098 -0.068 -0.271 0.038 -0.881 0.172 23 Staight Gluon. Technical comparison between JF and Mandy Inbetween Parameterisation Chi2_J = 442.3. Chi2_M = 437.9 Val_J Err_J Val_M Err_M Dif/Err Gluon Bg -0.091 Gluon Cg 11.433 Gluon Dg 11.341 u valence Buv 0.545 u valence Cuv 3.831 u valence Duv 3.496 d valence Cdv 5.042 d valence Ddv 6.006 Sea BDbar -0.210 Sea CDbar 4.958 Sea CDbar 3.895 0.028 1.026 3.970 0.047 0.107 1.028 0.605 3.279 0.004 1.722 0.851 Joël Feltesse -0.098 11.406 11.952 0.556 3.853 3.501 5.054 6.559 -0.209 4.615 4.615 0.034 1.310 5.463 0.058 0.112 1.230 0.713 4.585 0.005 0.869 0.869 0.242 0.023 -0.129 -0.211 -0.206 -0.004 -0.019 -0.490 -0.368 0.264 -0.837 24 Conclusion • The largest difference between JF and Mandy on fit results has been understood. • Technical comparison between JF and Mandy : agreement at the ~ 0.3 sigma level on parameter values for all choices of parameterisation, although Chi2 of Mandy are always a bit smaller. • At present, should concentrate on understanding (improving) the remaining differences between parameterisations and move to more elaborate fits, for example on HQ treatment and with treatment of correlations between errors. Joël Feltesse 25 APPENDIX Joël Feltesse 26 PDF parametrisation : x f(x) = A xB (1 - x)C (1 + D x + F x3 ) ZEUS-JET parametrisation (11 parameters) A From Sum Rule From Sum Rule gluon uv B C D F 0. 0. Ubar U From Sum Rule dv = Buv 0. Dbar D ubar - dbar Sea from from from 0. Z_S_11 fit Z_S_11 fit Z_S_11 fit 0. 0. 0. Quite simple but questionable assumption on ubar -dbar Joël Feltesse Where U = u +c and D = d+ 27 s+b PDF parametrisation : x f(x) = A xB (1 - x)C (1 + D x + F x3 ) H1 parametrisation (10 parameters) A From Sum Rule gluon B C D F 0. uv A(Ubar ) = = B (U) A(U) ubar/dbar→1 as x → 0. Ubar U 0. 0. From Sum Rule 0. dv A(Dbar ) = A(D) Dbar D = B(U) 0. 0. = B(U) From Sum Rule 0. ubar - dbar Sea Very strong assumptions on B’s, questionable assumption on ubar/dbar as x → 0. Joël Feltesse 28 PDF parametrisation : x f(x) = A xB (1 - x)C (1 + D x + F x3 ) In between (EP) 12 parameters A B From Sum Rule From Sum Rule ubar/dbar→1 = B (Dbar) as x → 0. gluon uv Ubar C D F 0. 0. 0. 0. U From Sum Rule dv = B (uv) 0. 0. Dbar 0. D ubar - dbar Sea Weaker assumptions on B’s. Less model dependence Joël Feltesse questionable assumption on ubar/dbar as x → 0. 29 EXTRAS Joël Feltesse 30 Joël Feltesse 31 Joël Feltesse 32 Joël Feltesse 33 Joël Feltesse 34
© Copyright 2026 Paperzz