| QCD Meeting ** Minutes ** |
||||||||
|
November 19, 1999 1:00 PM Pump Room |
||||||||
|
The presentations are available by clicking on the speaker.
|
||||||||
| Agenda: | ||||||||
 1. News/Announcements/Run II
|
||||||||
| Blessings: |
||||||||
| ||||||||
| Updates: |
||||||||
| ||||||||
| Minutes: | ||||||||
|
1.1 QCD Web Pages Webmaster Rick Field has agreed to revamp the QCD web pages; if you have any suggestions please send to him; if there are any blessed plots that need to be posted, please also contact him (as well as Anwar and Joey) 1.2 QCD Papers of the Week hep-ph/9911256, Hard Scattering in high-energy QCD, M. Mangano; summary talk on QCD given at QCD99 hep-ph/9911379, M. Kimber, A. Martin, M. Ryskin, Unintegrated parton distributions and prompt photon hadroproduction; consideration of direct photon production using pdf's with kT information still in; comparison to fixed target and Tevatron hep-ph/9911340, T. Binoth et al, A full next-to-leading order study of direct photon pair production in hadronic collisions; photon pairs from fixed target to Tevatron to LHC, including NLO fragmentation contributions 2. Diffractive J/psi production (Andrei Solodsky) Andrei is reblessing this analysis taking into account a correction and an improvement to the analysis. J/psi production takes place primarily through gg scattering and thus diffractive J/psi production probes the gluonic content of the Pomeron. The previous analysis used only single vertex J/psi events and the muons were required to be in the CTC. The ratio of diffractive J/psi to total J/psi (times the gap acceptance) was calculated as the ratio of diffractive J/psi to non-diffractive J/psi times an efficiency correction for requiring a single vertex times an efficiency correction for noise towers. There was an additional correction, based on instanteneous luminosity, to determine the number of single interaction events in the sample, resulting in a double correction. The resulting ratio times gap efficiency was found to be 0.64+/-0.12%. (See CDF note 4699.) With the double correction removed, the diffractive/total ratio times the the gap acceptance was found to be 0.37 +/-0.07%. This number was blessed. Andrei then described an improvement to the analysis. Both of the muon tracks from the J/psi were required to be in the SVX. This resulted in a cleaner J/psi sample; the resulting diffractive ratio was found to be 0.37+/-0.07%, which is very similar to the previous result. This sample also allows a separation of the prompt and B-decay components. The proper decay length, c-tau, was fitted for non-J/psi backgrounds, prompt J/psi's and J/psi from B-hadron decay. 95.5% of prompt J/psi are within c-tau of 105 microns while only 14.5\% of the J/psi's from B decay are within a c-tau of 105 microns. The fraction of prompt diffractive J/psi to prompt total was found to be 0.37+/-0.08%. This result was preblessed. For J/psi + jet events, we can measure the diffractive to non-diffractive ratio as a function of x_p(pbar). The result was found to look similar to the Roman pot dijet data, although with much larger statistical errors. Andrei will come back to a later QCD meeting to bless the new/improved portion of this analysis. 3. A few more MLLA plots (Alexei Safonov) In this continuation of the MLLA analysis, Alexei presented some fits for the MLLA parameters and some comparisons to Herwig. As a quick reminder, in the MLLA (modified leading logarithm approximation) formalism, the analytical results are infrared stable, with a cut-off parameter which is of the same order as Lambda QCD (or about 250 MeV). Hadronization occurs locally at the last stage: the hadrons remember features of the parton distributions and the number of hadrons/number of partons is a constant. In this analysis there are 9 different dijet mass bins, with the systematic errors being strongly correlated between the different bins. A chisquare calculation can be carried out taking into account the systematic errors, with the correlation coefficients being estimated and varied within a reasonable range. A comparison of the charged multiplicity as a function of the dijet mass for different cone sizes indicates good agreement with Herwig, if the Herwig charged multiplicity is scaled down by 11%. This rescaling was found to be constant throughout the fits for reasonable variations of the correlation coefficients. The hadron multiplicity can be written in terms of a constant (K_LPHD), the low energy cutoff Q_eff(240+/-40 MeV) and the term r. r is the ratio of multiplicity for gluon jets compared to quark jets. The result is K_LPHD= 0.69+/-0.03+/-0.05 and r=1.70+/-0.16 in agreement with Herwig and with expectations. The jet shape, as defined by the charge track momentum distribution transverse to the jet axis, is found to be in good agreement with the Herwig predictions, if the Herwig hadron multiplicity is rescaled by the factor of 0.89. Note that these data require two jets in the event which approximately balance and thus the effect of jet merging/splitting is reduced The results were blessed. A PRL on multiplicity is almost ready (and needs godparents); a PRD on momentum distributions is in the final stages of preparation and updates plots and fits will be presented for reblessing in the near future. The photon + jet analysis is underway and preliminary results may appear in a month. 4. A study of the underlying event in jet events (Joey Huston) This is the first time that this somewhat complicated analysis has been presented so this will be a long explanation. Sorry. Joey presented the results from an analysis conducted by Joey, Anwar Bhatti, Eve Kovacs and Valeria Tano, with Valeria having done most of the work. Valeria is a student at Munich who was/is a visitor to CDF through Michigan State. The paradigm that has been adopted (for both CDF and D0) for comparison of jet cross sections to NLO QCD is that a jet event consists of two (or sometimes three) jets sitting on top of a "minimum bias" underlying event. The hard interaction may have any number of jets but we usually compare it to NLO calculation which has at most three jets. Thus the minimum bias level of energy needs to be subtracted from the jet event before any comparisons are made. CDF subtracts the energy in "active" (class 12 vertex) MB events and assumes a 30% uncertainty in the amount to be subtracted. This is the largest systematic uncertainty for CDF in the ET range below 60-70 GeV. Note that D0 assumes no uncertainty for this subtraction. Previously, for leading order analyses, the energy in 90 degree cones was subtracted; this is not the procedure to follow for NLO analyses since this may subtract energy already taken into account in the calculation. The 1989 inclusive jet PRL and XT analysis used a 90 degree cone energy subtraction. Both are NLO analyses. It was changed because the theorists threatened to beat us up if we didn't. The current procedure may potentially ignore a number of other effects such as double parton scattering, higher order radiation ("splash-in"). See, for example, the paper by Jon Pumplin, PRD57 (1998)5787. A better understanding of this subtraction will lead to significantly reduced errors for the low ET jet cross section, and a better understanding of low ET jets, period. This may have impact on the 630 GeV jet cross section and the xT scaling problem, since the effect is primarily for low ET jets (and the xT scaling problem derives from the disagreement of the 630 GeV low ET jet cross section with theory). The data analysis used jet ntuples from Anwar that had information on the tower energies in the central rapidity region. A requirement was made that the data have to have 1 class 12 vertex and no other class 10,11 or 12 vertex. Comparisons were made primarily to Herwig (version 5.6) with the generated Herwig events passed through QFL (although some comparisons were made on the parton and hadron level). Data samples were constructed for both the data and Herwig corresponding to JET20, JET50, JET70 and JET100. For each event where there was a lead jet in the central region (the lead jet had to be in the rapidity range of +/-0.7, similar to the inclusive jet analysis) two cones of radius 0.7 were examined, at the same rapidity of the lead jet but +/- 90 degrees in phi. The energy in these cones was determined for each event: two tower thresholds were considered, 50 MeV and 100 MeV, with the 50 MeV threshold being default for most of the following. Also calculated was a 'Swiss cheese' energy consisting of the ET in the central region (+/-1) once the two (or sometimes 3) leading jets were cut out. The 90 degree cone that had the most energy was called the 'max' cone and the other the 'min' cone. As a function of the lead jet ET, both the Herwig and the data have the same features. The max cone increases with lead jet ET, peaking broadly at about jet ET's of about 150-200 GeV, with perhaps some indication of a drop at the highest jet ET's. For the data, the max cone starts off at a value of about 3 GeV and peaks at about 4.5 GeV. The min cone lies at about 1.1 GeV. The min cone is almost perfectly flat as a function of the jet ET. It is natural that the max cone increases with jet ET, and the plateauing and drop at high ET are probably just due to the kinematic restrictions at high x. One might have imagined, though, that the min cone would also tend to increase with jet ET, although perhaps not as fast as the max cone. Unlike the 'Republican mantra', a rising tide does not raise all boats in this case. The main difference between data and Herwig is the offset observed for both the max and min cones (data greater than Herwig by about a constant 800 MeV for the max cone and 500 MeV for the min cone). A plot of max-min for each event shows Herwig and data to be very similar except for the remaining few hundred MeV offset. The min cone is about 200 MeV above the ET level observed in min bias events, for both Herwig and the data.Note that this means that the Herwig min bias event is very soft compared to the data. The max-min distribution should be a measure of the NLO (3rd parton) contribution to the event. Sometimes the 3rd parton will land in or near the max cone; any partons landing in or near the min cone are beyond NLO. One question that remains to be answered is why the Herwig underlying event is less than the underlying event observed in the data. It could be due to problems with QFL simulation for low pT particles, or it could be due to problems with the Herwig simulation of the underlying event. In Herwig the underlying event is assumed to be a soft collision between the two beam clusters, containing the spectators from the incoming hadrons. The UE model comes from the ppbar event generator from the UA5 experiment, scaled up to Tevatron energies. The same algorithm is used to simulate MB events. A random cone study in MB events finds that the average ET in Herwig (.37 GeV) is about 500 MeV less than that observed in the data (.86 GeV). Perhaps some additional tuning or the presence of double parton scattering is necessary. The authors have been in contact with Bryan Webber and have tried some of his suggestions at modification, but nothing that is reasonable has a significant effect (but DPS has not yet been tried). A comparison was made of the frequency distributions, for the max, min and max-min cones for Herwig and the data. The max-min distributions look very similar for Herwig and the data. These results were for a 50 MeV tower threshold. The result is very similar if 100 MeV tower threshold is used. The phase space inside the central region is very tight. As a check, two additional cones, at the negative of the rapidity of the lead jet and again +/- 90 degrees in phi were defined. An additional requirement was also made that the two most energetic jets had to be on the same side of the detector. For this situation, the average energy decreases by 50 MeV for the min cone and 100 MeV for the max cone. The Swiss cheese results show a similar behavior to the max cone; if the leading two jets are subtracted, the Swiss cheese energy (data) starts at about 18 GeV for jet ET's of 40 GeV increasing to about 30 GeV for jet ET values of 150-200 GeV with perhaps some indication of a decline at the highest ET values. If the three leading jets are subtracted, then the distribution is flatter and at a lower energy, varying from about 14 to 17 GeV. Again, Herwig(+QFL) shows a similar behavior to the data, albeit with an offset of 3-4 GeV. Note that if the energies of the 3 leading jets are subtracted, then there can be no 'NLO contribution' to the energy in the central region. The difference between the Swiss cheese energy with 3 jets subtracted and the estimated min bias energy in the central region varies between 6-9 GeV (at the detector level). Other contributions to the Swiss cheese energy could include higher order radiation and splashout. Some tests were conducted with Herwig turning resonance decays off; the result was that the Swiss cheese energy was smaller at the hadron level (closer to the parton level, since fewer hadrons are "splashing out" of the jet cones into the Swiss cheese). The difference was about 2-3 GeV at the hadron level (probably divide by about 1.5 to get to the detector level), not enough to explain the difference between the Swiss cheese-3 jet subtracted distribution and the expectation from minimum bias. Another approach to this question of NLO comparisons is to compare the predictions of JETRAD for the max/min cones and for the Swiss cheese energies. In events with 3 partons in the final state, the max and min cones can be defined in the same way for JETRAD as for the data and Herwig. Of course, the min cone is always zero, as is often the max cone. The same type of initial increase and then plateauing is observed for the JETRAD max cone distribution as was observed for the data and Herwig (although the plateauing seems to take place at a higher ET value. Definite predictions may be hard to come by, both due to the uncertainty of the parton->detector scale and the sensitivity of the calculation to the renormalization/factorization scale in the calculation. In fact, the scale sensitivity of the JETRAD predictions is almost alpha_s^3, due to the numerator in the calculation being LO alpha_s^3 while the denominator is NLO alpha_s^2. The denominator has little scale dependence, allowing the full LO scale dependence of the numerator to come through. Similar plots were shown for the Swiss cheese calculation. The appropriate scale to evaluate the 3rd parton (the one that contributes to the max cone) probably is the ET of the 3rd parton (which was the original Marchesini-Webber suggestion). The JETRAD events can be rescaled by alpha_s(ET_3)/alpha_s(ET_max). Another possibility (in addition) is to make the denominator in the calculation of the max cone energy LO instead of NLO. This will cancel some of the large scale dependence. The goal is to bless at least some of the plots in the near future. Some of the projects (besides the ones mentioned in the above paragraph) include considering the uncertainties for QFL simulation of low energy particles, and to compare tracking information in the underlying event samples to Herwig (where we can be independent of the calorimeter energy scale for the low energy particles). 5. NLO 3 jet production (Alex Brandl) Alex presented an update on the study that he and Sally Seidel are doing regarding the measurement of alpha_s by the comparison of energy partititioning in multi-jet events to predictions from a NLO calculation. The NLO 3 jet calculation was performed by Bill Kilgore and Walter Giele (hep-ph/990-3361); the data used in the comparison came from the CDF sum ET stream B data banks (sum ET > 175 GeV, 86.5 pb^-1). The analysis uses systems of three or more massless jets. The three leading jets in the laboratory frame are used as the basis of transformation into the three-jet rest frame. In the rest frame, the jets are re-ordered according to their energies Ei. Five variables are needed to fully characterize a three-jet system. The variables customarily used are the mass of the three-jet system, the dimensionless Dalitz variables x3 and x4 (xi=2Ei/m3j) and the two angular variables cos(theta*) and psi. To remove beam halo, calorimeter malfunctions and cosmic rays, the following cuts were applied:
The jet correction JTC96S was applied to the data (underlying event correction applied, out of cone correction not used). Three or more jets were required for each event with a minimum jet ET cut of 20 GeV and a requirement that the rapidity of each jet be less than +/- 2. A cone size of 0.7 was used. An angular cut on cos(theta*) was applied to ensure that the geometric efficiency was greater than 95%. The Monte Carlo prediction involves one loop 2->3 contributions and tree level 2->4 contributions. The Monte Carlo output consists of 2 parts, a hard emission with all partons well resolved and an infrared part with one or more partons in the soft or collinear limit. The two parts are added algebraically. There are theoretical errors due to the limited Monte Carlo statistics and to neglect of higher order terms. For the analysis, 10 template Monte Carlo data sets were generated, each one equivalent to 86 pb^-1. Each template was made with a different value of alpha_s, using the appropriate parton distribution function in the CTEQ3M family. The Monte Carlo data sets were then compared to the data, with the conjecture that the variation of the density in the Dalitz plane provides a measure for alpha_s. The data are binned in x3-x4 space (0.02X0.02) and the binned Monte Carlo cross sections are multiplied by the luminosity for the data set to obtain the expected number of events in each bin. The data are compared to Monte Carlo using the method of maximum likelihood. No unsmearing has been done yet. The chisquare is plotted versus alpha_s (corresponding to each pdf) and a minimum in the chisquare is sought. A (weak) minimum is found at an alpha_s value of about 0.115. The theory (statistical) errors are still large due to only two weeks of running to obtain the cross sections (6 weeks is recommended). The same exercise performed with Herwig obtains back a similar alpha_s to that put in. Things in progress:
A chisquare can then be calculated which measures the bin by bin difference between the measured number of events in a bin and the expected number, given a "true spectrum" and the given response function. From the minimization of this chisquare, the parameters of the true distribution can be determined. A correction factor can then be defined for each bin as the smeared value divided by the true value.The data can then be divided by this correction factor (bin by bin) to give unsmeared fully corrected values. The next steps:
6. Use of theory errors in chisquare calculations (Eve Kovacs) First, some background to the story. D0 has recently released their inclusive jet cross section for all five rapidity intervals (up to a rapidity of 3.0), with a total of 90 or so data points. The best visual agreement is obtained with cteq4hj which also gives a nominally better chisquare than other pdf's such as the MRST series. The chisquare was at first calculated using systematic errors computed using the experimental cross section, similar to the technique that CDF has been using. When D0 tried calculating the systematic errors using the theory for normalization instead of the data, the result was a great chisquare for cteq4hj (as before) with the chisquares from the MRST series becoming significantly larger. No other pdf has a "snowball's chance in hell" of describing the D0 data. The rumor is that D0 switched its procedure due to a strong lobbying effort by Walter Giele and Stephane Keller. Eve discussed the adoption of this procedure for CDF as well. The motivation is that we want to know the probability that our measurement resulted from fluctuations of a given theory distribution. Fractional systematic errors computed from the STANDARD FIT to data multiplied by theory would give a good approximation to the exact result, since the fractional errors depend only on the shape and not on the normalization. Eve showed some results for the chisquare calculations for the dijet mass cross section for several pdf's. There is little difference in the chisquare for cteq4hj if the theory is used to calculate the systematic errors rather than the data (because the normalization is about right) while there is a large increase in the chisquare for MRST3 (where the normalization is very low). Note that the chisquares become worse for the MRST series since their normalization is low. If this procedure were used for theories with a larger normalization than the data (of which there are currently none), the result would be larger errors than if the calculation were done only with the experimental data. This procedure will be adopted for both the dijet mass and differential dijet cross section. 7. Plans for diphoton analysis (Bob Blair) Bob discussed plans for publication of the diphoton analysis from Run 1B. The results have been analyzed for some time but publication was delayed after the graduate student (Takeshi Takano) primarily responsible left the field. The diphoton fraction in the Run 1B sample has been determined using the standard CES/CPR weighting techniques, complicated by the fact that there are two photons rather than just one. The backgrounds to the diphoton signal come from photon-pi0 and pi0-pi0 events. The photon fraction can be determined with the inversion of a 4X4 matrix. The sources of systematic errors are from the uncertainty in the CES and CPR background subtractions, the uncertainty in the L2 trigger efficiency, the uncertainty in the isolation cut, the uncertainty in the no-tracks requirement, the uncertainty in the photon energy scale and the luminosity uncertainty. The diphoton cross section has been plotted versus a number of kinematic variables including photon pT, diphoton mass and diphoton transverse momentum. The latter distribution, in particular, shows the need for resummation of soft gluon effects in order to have a proper description. Things to do:
|
Joey Huston - December 4, 1999