Muon Offline Meeting -- Minutes ================================= 30 - Oct - 2002 ---------------------------------------------------------------------- Camille Ginsburg -- BMU Z analysis and silicon-only tracking ======================================== Camille has started looking at the the high-pT muons that have recently been reprocessed with 4.8.4, getting through about 48 pb-1. She relinked her muons to allow SI-only tracks, to see if she could increase the BMU yield that way. She makes pretty loose cuts on the muons, requiring one in the CMUP and one in the BMU, with minimum-ionizing cuts and |dx| < 10 cm for both muons (but she wasn't sure whether it was CMU or CMP or both for the dx cut). She gets 135 Z's as a result, ten of which have an SI-only track in the BMU. (Note that comparing this work to her last pass, she has 72 Z's in the run range she looked at before, compared to 63 in that last pass. The nine new ones all used to be called cosmic rays!) She finds that the beam constraint is important for improving the width of the peak -- it goes from 7.4 to 5.6 GeV with the constraint. The SI-only tracks tend to be at higher eta than the other BMU tracks, which makes sense. Otherwise, they have flat spatial distributions. In short, this is all good news for both silicon tracking and the BMU! Ten isn't a large number, but we should be happy. Camille also noted what she called the "Pondrom curiosity." She and Lee calculate dz for BMU muons on the fly, using the track eta and z0 to estimate the z of the track at the BMU. The higher-eta muons (have energy in PHA rather than WHA) show a dz shift that is shifted 30 cm off of zero. A bias in z that is z-dependent? She should look at dz vs. zstub or ztrack, rather than just breaking it into two bins. Jim Bellinger -- Recent BMU reconstruction advances ================================== Jim has been studying how using the crossing time, i.e. the time that a particle crosses the BMU, can be used to improve the stub finding. By using scintillator hits, he can clean up this distribution, and see components from real collisions, beam halo, and splashes. This allows him to focus on the prompt stuff. These stubs have a noticeable width in the crossing-time distribution, about 4 ns, and that can be used as a correction in the stub finding. This then starts to play into some alignment issues. Chisquared distributions for different kinds of stubs (e.g. three-hitters, those that cross stacks) look different, and that suggests alignment problems. The Bob Handler stub finder is also brought into play for these studies. Putting an improved alignment together with the crossing-time corrections leads to lower chisquareds, and ultimately fewer and better stubs. Jim saw that his muons do not seem to be evenly distributed -- there is a deficit on the west side, in a certain eta-phi range. This needs study. He is also working on incorporating scintillators into the reconstruction, for further cleanup. Slava Krutelyov -- Retuning the MExtrapolator for the CMX: data ============================================ Before Slava looked at the data, he took another look at the simulation, beating the MExtrapolator with improved tuning against the G3X extrapolator, which ought to reproduce the simulation very well, since the simulation uses GEANT. Strangely, G3X doesn't work quite as well -- for both positive and negative muons in the CMX, the dx distribution has a mean at a few mm less than zero. Meanwhile, the MExtrapolator tuning is good to about 0.5 cm at 2 GeV/c, and twice as good above that. However, there are effects in certain regions of phi even at high pT. Slava believes that this is an effect of how GEANT handles empty space -- it takes big steps when there isn't much material, even if the B field is changing rapidly. By contrast, he makes small steps through a changing B field. In addition, the MExtrapolator shows disagreement at the smaller eta parts of the CMX, where the various support structures are, where the CHA becomes the WHA, and where the magnetic field is turning over in the yoke. (In short, we are running into all the hard parts of dealing with geometry which GEANT should handle more easily! Should, at least.) As for the data, Slava looks at J/psi muons (but with no sideband subtraction, which might have an effect). Is this the first time that anyone has looked at the dx at small pT from real muons, separating out the two charges? Even in the CMU, the dx mean deviates from zero by about 1 cm at 1.5 GeV/c, showing the classic 1/pT behavior. [Ed. note: I asked Tom LeCompte to look into this, with a sideband subtraction, and he confirmed the effect in the CMU.] It sounds scary, but we should note that the typical multiple scattering for such a muon is 10 cm, so a 1 cm shift is not so big within the reasonable cuts that someone would make. Jonathan (who showed up at the meeting by accident, it seems) pointed out that this is probably related to an inferior estimate of the flux in the CHA, due to our ancient and imprecise field map. Results in other detectors were harder to understand because of their more dificult geometries; in particular, the CMX seems to have dx offsets of order 1 cm regardless of pT (within the extent of the J/psi pT spectrum, which goes up to 10 GeV/c). It looks like we are tuned OK to simulation, but the tuning to data doesn't look so hot. Will we have to take on a program to understand the field map? If we are worried about things at the 1 cm level, maybe, but considering typical scattering at low pT, it's probably not a top priority. Petra Merkel -- High-pT muon fake rates ======================= Petra and Ray Culbertson and others have been looking into fake rates from punchthrough and decay in flight, via three different methods. The first (and most entertaining, in my opinion) was through K_S and D0 reconstruction, which gives a clean source of pions and kaons. They use the two-track trigger to get a very clean sample, with very small sidebands for the K_S (not as small for D0), and good statistics even up to 10 GeV/c. (There's a 2 GeV/c cut in the trigger.) They use ICHEP-like cuts to identify muons (but no isolation cuts), and match these up with the pion and kaon tracks from the resonances to see how often they are identified as muons. The fake rate for a pion from K_S to show up in any detector is measured as (0.38 +- 0.02)% over all eta, no fiducial cuts. Most of this fake rate shows up in the CMU, of course. From the D0, the pi to mu fake rate is (0.46 +- 0.03)%, and the kaon fake rate is (0.75 +- 0.03)%. The difference in pion fake rates might be due to the different momentum spectrum. They do this measurement in other trigger samples too (e.g. photons, electrons) and find somewhat lower fake rates. Is there a physics bias in the samples? There might be more real muons in the two-track sample due to real semileptonic decays. To study this, they varied the matching cuts; loosening them brings in more apparent fakes, of course, and this effect seems to be bigger in the two-track sample. If these fake rates are compared to Jeff Berryhill's work in Run 1, the background estimates for his analysis come out to be roughly the same. The second method uses tracks from inclusive samples, which are more abundant, and gets you to higher momentum, but you can't say you know the original particle identity as well. The results are consistent. In jet samples, the fake rates are lower if the track is the middle of a jet. The hypothesis is that the minimum-ionizing cut tends to take you out of the jet, and that jets with muons in them have less energy, and are less likely to make it in the sample. In general, there is good agreement with all samples, and with simulation, which is the third method. There might be some trouble with cosmic contamination, seen as apparently much higher fake rates for isolated tracks. Petra and company are making refinements, and hope to make another presentation again soon. Michael Gold -- CMP resolution in data and simulation ===================================== Ken actually presented Michael's one slide, which had fits of resolutions in the CMU, CMP, and CMX. The observed CMP resolution is about 0.18 cm, which is much larger than the 0.025 cm currently in the simulation, and Michael would like to change the latter to the former. Should we? We agreed that there will be little impact on anyone's analysis because of this, so why not? But the hardware folks were a bit puzzled, and want to think about it. Hyunsoo will get in touch with Michael. Victoria Martin -- Dead-channel database ===================== Thanks to Victoria, Camille and Phil we will soon have a functioning channel-status database. For each channel, there will be a word indicating the status of the channel through various bits, including bits for whether it's OK or not for analysis. Victoria will make easy accessors for C++ code, e.g. status->dead(code). The database would be updated when there is enough data to make a judgement call about a channel, or when there are physical changes in the detector. How might we want to use this for reconstruction? Exclude noisy channels from reconstruction? It's not clear that the noise bits will be updated carefully. Allow two-hit stubs in the CMU and CMP if there's a dead channel around? Maybe, but if the channels are randomly scattered, this will only be an issue if you are unlucky and lose one of the three remaining hits. Phil pointed out that with the HV status bit, we might start fixing wedges with no cathodes, or maybe that can just be done with calibration constants. We'll see! ---------------------------------------------------------------------- K. Bloom