******************************************* * Minutes of the EW task force (2-may) * ******************************************* Gerardo did some more studies on the discrepancy between present scanbook luminosities and the published papers for what concerns 90 and 91 data. Here I append the relevant information from a mail that he sent: 1. Sources used for the study ------------------------------- o 1. Table 3 in Z.Phys. C53(1992), "Improved measurement ..." o 2. Table 1 in Z.Phys. C60(1993), "Update of EW ..." o 3. New SCANBOOK output o 4. table of luminosities used in our fits (from Monica ...) o 5. Old SCANBOOK outputs from 90 and 91 (from B.Gobbo) o 6. List of runs used to measure tau x-sect in 89 and 90 (ALEPH 90-189) o 7. Luminosities used for tau x-sect 89-91 (ALEPH 92-023) 0. Problem ----------- Three different sources quote the following integrated luminosities for 90 and 91 (in nb**-1): 1990 ---- sqrt(s)-m_Z scanbook 96 paper 93 fit tables -3 482.0 480.2 480.2 -2 520.1 520.3 520.4 -1 447.2 444.0 444.0 ----> 0 3624.0 3504.6 3632.7 +1 554.6 553.8 553.9 +2 597.2 594.2 594.2 +3 641.8 641.6 641.7 1991 ---- sqrt(s)-m_Z scanbook 96 paper 93 fit tables -3 668.4 671.0 670.6 -2 796.8 798.9 798.4 -1 753.2 748.5 748.2 0 7546.4 7547.7 7547.7 +1 693.2 694.3 693.8 +2 677.3 680.0 679.6 ----> +3 796.6 765.0 764.6 The arrows indicate the data sets for which the discrepancies are larger. 1.1. 1990 --------- With the help of the run lists found in [5,6], I've found that the run selection given by scanbook is the same used at that time, except for 2 runs at the peak that now are excluded: 8506 ( run_quality: MAYB lumi: 4.735 nb**-1 ) 9070 ( run_quality: MAYB lumi: 23.058 nb**-1 ) Subtracting these two runs from the old run list gives at the peak 3627.8 nb**-1. This is in rough agreement with the scanbook and fit-tables values, but not with what found in [2]. In the paper 92 [1] the value used for the data set collected at peak in 1990, is 3655.7 nb**-1; this value contains (probably) also runs 8506 and 9070. 1.2. 1991 --------- The difference at pk+3 is due to run 12558 (run_quality: PERF, lumi= 28.385 nb**-1 ) which has been added now (recovered during reprocessing in 93?) Action-> Can Pere check this? 2. Status --------- a) the 1991 run selection given by scanbook is ok; b) 1990. Run selections out of peak are ok; at peak it looks like some runs were analysed for the 92 paper, disappeared for the 93 paper, and appeared again for the present fits ... In the 93 paper, it is shortly mentioned that "because of the use of relative luminosity for 1990 data, ..., new values are given for the 1990 cross sections superseding the previous ones". However I did not succeed so far to find trace (in aleph-notes) of this "temporary" change in the run selection. I think it is important to understand exactly what happened so we can be sure of what we are doing now ... ****Could maybe Tom Fearnley or Alain help understanding this point?**** c) in all cases luminosities did slightly change ... ************************************************************************* Andrea Venturi pointed out that run 20766 from 93 appears when running scanbook and asking for the EW selection, but in reality the DSTs do not exist for this run. Do the POTs exist? Should that run be removed from the EW selection? Action ---> A.Venturi, P.Comas, L.Thompson. After the meeting Pere clarified that "run 20607 could not be reprocessed (irrecoverable crash in JULIA) and is therefo- re missing in this reprocessing. This is a MAYB run with 538 hadronic Z0s" (from Jacques B. news on the end of the reprocessing of 1993 data). Is therefore suggested to take out this run from the EW selection of 1993 and pos- sibly to DUCK it. ************************************************************************* Marie-Noelle pointed out that the reprocessing of 90-91-92 data took place in july-oct.'93, while for ex. 93 data were reprocessed in sep.95. Are we sure that that old reprocessing is OK? --> Action Pere and Andrea ************************************************************************* Monica showed the global statistics appearing in Scanbook for the EW and EA selections: EW EA ( nb. qqbar ) 90 146.933 91 286.371 92 674.726 93 635.351 94 1.269.726 1.639.854 95 361.438 736.197 Total 3.374.545 4.119.432 Difference = 744.887 So if we measure Rl only for the EW selection, we loose 22% of the data, which is clearly unacceptable. Therefore we decide that since for Rl we do not need the luminosity, for that measurement we will use the EA selection. People doing the analysis should therefore provide numbers for both the EW selection (cross-sections) and the EA one (Rl). ************************************************************************* Bolek clarified what radiative corrections are present in the different versions of KORALZ. Up to 93 in Aleph we used KORALZ 3.8 (ALEPH 6) which is a version where only initial state exponentiated and single final state bremsstrahlung are generated. From '93 we moved to KORALZ 4.0 (ALEPH 7) containing initial and final state bremsstrahlung O(alpha**2) with exponentiation a' la YFS2/YFS3. Therefore multiphoton final state radiation is introduced (but QED initial-final state brems. interference in the presence of multiple hard bremsstrahlung is not included. (more in CERN-TH 7075/93)). Status of the generation in ALEPH: 1. 100k mumu were generated with version 4.0 of KORALZ, setting KEYRAD=112 to be compatible with KORALZ 3.8. With this setting only ONE final state photon is generated (run 911). 2. Another 100K were generated with KEYRAD=12, allowing MULTIPLE final state production (run 1757). These statements were supported by plots by A.Venturi on the number of generated photons. In addition we have the MC generator by Miquel and Martinez which generates mumugg events at the 4-momentum level (exact O(alpha**2) calculation). The comparison of these various codes should allow to reassess the error associated to higher order radiative corrections. Henri reccomends not to use MC generated with versions 3 and 4 (ALEPH nomenclature, see minutes of last meeting). ************************************************************************ Status of UNIBAB. Ioana raised some questions which were answered by Lee and discussed at the meeting. 1) Do we have the latest UNIBAB version? We have a semi-official version (1.99/08) that is not mentioned in the manual (probably due to Helge's interest in the generator). Lee thinks that our version is essentially the same as 2.0. In this hypothesis the differences are: Version 2.1: beam polarization introduced Version 2.2: update of the EW library Improved final state photon shower algorithm different choices of factorization scale for FSR QED virtual corrections Speed-up of EW corrections Fixed minor bug in initial state photon shower Lee has mailed the principle author, Anlauf, to ask for his comments on the importance of these features and to check exactly what is in version 1.99/08. 2) Is the luminous region defined in the MC for each year? Answer by Brigitte: The luminous region is steered using data cards. 3) Which acceptance should we generate? UNIBAB ONLY accepts COSTHETA* cuts (contrary to what said at the meeting). Lee suggests to generate events with |costheta*|<0.91 and Acoll<30 degrees. Some people thought that we should not cut in the acollinearity, since that is not a necessary cut. -> Lee, Ioana please react. 4) Lee proposes to start generating 94/95 geometry at P, P+-2 with the current version of Julia. Then we should continue with 90/91 geometry P+-1, P+-3. This prod. should start asap and Lee should split the work between Monica, Ioana and himself --> action, please. Others are also waiting for this production.. ************************************************************************ Lee made a study to see the effect on the statistical and systematic errors of varying the angular range over which the cross-section is calculated and the asymmetry fitted. He took 1993 data which is well understood. The assumptions is that the systematic errors are the same as always, in particular a 2% error on the t-channel subtraction cross-section is assumed. In the figures you see below the only changes that are made are 1) the angular range and 2) the acceptance correction which is calculated using 1 million mu pairs at each energy point. For the cross-section the errors quoted are in percent, for the asymmetries the errors are absolute. -0.9 to +0.7 ============ E Sigma Stat(%) Syst(%) AFB Stat Syst 89.431 493.653 +- 2.354 +- 1.406 -0.178 +- 0.030 +- 0.016 91.296 1498.660 +- 1.391 +- 0.465 -0.009 +- 0.015 +- 0.003 91.191 1466.680 +- 1.084 +- 0.489 0.021 +- 0.012 +- 0.003 93.017 699.687 +- 1.537 +- 0.385 0.131 +- 0.016 +- 0.001 -0.9 to +0.6 ============ E Sigma Stat(%) Syst(%) AFB Stat Syst 89.431 494.486 +- 2.282 +- 1.053 -0.177 +- 0.032 +- 0.014 91.296 1500.372 +- 1.421 +- 0.426 -0.009 +- 0.017 +- 0.002 91.191 1464.911 +- 1.107 +- 0.441 0.022 +- 0.013 +- 0.003 93.017 691.318 +- 1.597 +- 0.374 0.119 +- 0.017 +- 0.001 -0.9 to +0.5 ============ E Sigma Stat(%) Syst(%) AFB Stat Syst 89.431 493.041 +- 2.259 +- 0.843 -0.185 +- 0.036 +- 0.012 91.296 1505.361 +- 1.460 +- 0.409 -0.007 +- 0.019 +- 0.002 91.191 1464.990 +- 1.138 +- 0.419 0.025 +- 0.014 +- 0.002 93.017 687.946 +- 1.668 +- 0.374 0.113 +- 0.019 +- 0.001 No clear conclusions emerged from these numbers which were received just before the meeting. They will be rediscussed next time. Next meeting: 9 may at 15:30 room 1-1-045.