From kenbloom@umich.edu Fri Jun 14 21:59:08 2002 Date: Tue, 14 May 2002 18:04:06 -0400 (EDT) From: Ken Bloom To: cdf-muon-offline@fnal.gov Subject: meeting minutes for 5/1/02 Muon Offline Meeting -- Minutes ================================= 1 - May - 2002 ---------------------------------------------------------------------- Ken Bloom -- Electron-muon common issues =========================== We opened in joint session with the ETF to talk about the things that both groups need to get done for the summer. The top/ewk group has asked our groups to produce W cross section measurements on that timescale. We only have eight weeks left before a result for ICHEP must be preblessed, and there is a lot to get figured out by then. This was meant to be an opening discussion of these issues; it will not be the last one. We would hope to establish some common solutions and policies, but of course that is not required. As the minute-taker was also the speaker, it is a little hard to reproduce the discussion; there will be some gaps in this narrative. Do the two groups even want to analyze the same set of runs? The answer appeared to be "no," as different parts of the detector were working at different times. We're really not interested in muons from before 2/9, while the calorimeter and its triggers were working better earlier. We expect to keep an open mind about the last date that we can acquire data for a summer analysis; going up to the June shutdown is possible. But figuring out the good runs during this period will be the hard part; we are clearly going to have to go through the history of things quite carefully. As just one example, the electron group will have to account for the period when the Level 3 trigger was buggy. (And we'll probably all have to account for the fact that the COT tracking was throwing out high multiplicity events.) As for reprocessing, it turns out that the top/ewk convenors have already requested that Stream B be reprocessed with 4.5.0. It seemed generally agreed that this will be the offline release that we will use, rather than waiting for a 4.6.0. (That release will have improved silicon alignment, but that's not important for us.) People expressed satisfaction with the current state of the luminosity accounting, and dissatisfaction with the primary vertex finding efficiency. Ashutosh Kotwal -- Cosmic rays with COT timing =========================== Cosmic rays are out of time in two senses -- t0 is shifted, and one of the two legs is going the wrong way, outside-in, so the speed of light has the wrong sign. For a real track, the hits are later as you go from inside to outside, the opposite for an inbound track. So, for each track, float the t0 and beta^-1 and do a chisquared minimization. In Stream A data, Ashutosh makes a diumon selection, requires them to have z0 close together; these should be either real collider dimuons and cosmics. He selects collider muons by selecting Z mass, small impact parameter and not back-to-back. Cosmics are selected with large impact parameter and very back to back. He gets a reasonable Z peak, and the cosmics provide a falling background in invariant mass. The changes in beta and t0 from the refit are small for the Z muons, and bigger for the cosmic muons; in fact, they barely change at all for the collider muons. Ashutosh thus tries a simple box cut in the t0-beta^-1 plane, such that the outgoing track will fail the cut. The efficiency of the cuts is at least 90%, with a rejection factor of about 220. The two-dimensional cut is better than doing just one or the other. This seems an encouraging start. Anyes has the code and is using it in the context of the cosmic-ray tagging module. Ashutosh intends to work soon on the problem of an entirely missing track, which this technique does not address. Henry offered some more ideas, involving timing differences between the superlayers for both the incoming and outgoing track; the outgoing track helps to set the t0 for the incoming. This was developed in Run 1, and could be of good use here. Ashutosh says he has ben considering similar ideas in somewhat different language, and that this sort of plan will help for the missing leg. Will these techniques work at high eta? Ashutosh says that you need 3 or 4 superlayers, but it should work fine with that many. Chris Hays -- Muons in top/ewk validation =========================== Chris is trying to do the same things in his module with muons as with the electrons, with the most important thing being to look at quantities as a function of run over time. We agreed that there is overlap between different validation modules in terms of what they are meant to monitor, but that it's OK for now. He looked at Stream A data covering from sometime in the fall through early March, so some of the data are in fact very old. His cuts for identifying muons are somewhat different than what is being used by the group, and he was encouraged to make them more similar. In particular, CMU muons used, rather than CMUP (which is what the trigger is). He does separate out cosmics using the CosmicFinder. More work will be needed to make the true top/ewk part of the monitoring (i.e. W and Z rates) more useful. Chris sees differences over time, reflecting the changes in the reconstruction code and running conditions over time, or so we think; it was hard to tell given our poor memories. For instance, we could see differences that occurred when we switched the trigger from CMU to CMUP -- doing so greatly increased the muon purity. The plots for cosmics do look like those for real muons should. Chris wants to add more information, such as relative rates, information from monitor triggers, and more from CMX. Anyes Taffard -- W and Z samples =============== Most of this talk can be seen at http://www-cdf.fnal.gov/internal/people/links/AnyesTaffard/cdfsoft/stripWZ.html, which is Anyes's Web page describing what she did for the skims. She reprocessed just about everything we have starting from 1/14/02, using code that resembles 4.5.0, but isn't exactly it, for various reasons. It should be noted that running the production executable on that much data on fcdfsgi2 was quite a valiant effort! The samples are organized into different running periods and different streams, i.e. A and B. She has separated out the W and Z events according to the specifications that we had agreed on before. She also has all the log files and the luminosity files too. As agreed, the missing E_T for the W sample is calculated in three different ways -- the default with z=0, using the muon z0 as the vertex, and using a primary vertex. The Z events basically have two muon-like objects, one of which can be a stubless muon. Cosmics are tagged but not removed from the samples. Overall, there is about 19% cosmic contamination, 4% W, and 2% Z. Hmm, that's a strange W/Z ratio -- shouldn't it be more like 10 to 1? But further analysis with real cuts it comes out about right. Anyes also showed a few plots that she had just pulled out of the printer, and were not quite understood yet. A check of the dx distributions, which should now have the right alignment in, shows that the CMP still looks funny, with a long tail on one side. She'll look at it. Anyes also tried making the mass plots with beam-constrained fits, and does the Z peak have a funny shape? Look at changes and curvature before and after refit, looks like there are a few funny tracks. But this was all preliminary. Un-ki Yang -- One plot on trigger rates ========================= Un-ki looked at the trigger rates as a function of time, to make sure that we don't have a trap like in the electrons. He measures trigger cross sections from 2/10 to last Friday in Stream A. It appears that there is a drop in rate sometime around 3/23, and the scatter becomes much smaller (is that good?). A home-brew good-run requirement was imposed on the runs studied, makes sure that the all of the COT and central calorimeter appear to be on. So what changed? Nothing in the hardware, as far as we know. Best guess is a patch to Level 3, but we don't know what it is. Or maybe a change in the luminosity calculation? Agreed that we really need to understand how big the change is, and that it would be best to compare this to some other triggers and event rates, such as W --> e nu. Ken Bloom -- "Two" studies of muon efficiency ================================ Ken meant to show two studies, but we ran out of time, so he showed one and spoke very quickly about the other. Now it is time to start thinking about the denominator, and that means focusing on the efficiency. What is the efficiency for a real muon to be reconstructed as a CdfMuon? And then what is the efficiency for a real CdfMuon to pass identification cuts? In trying to answer both of these questions, Ken had to face up to the problem that the denominator in the efficiency measurement was itself uncertain, and errors had to be handled appropriately. To study reconstruction efficiency, Ken looked at a sample that had 4 GeV/c CMUP muons plus an SVT track. He chose one good CMUP muon as the potential first leg of a J/psi, and then looked for tracks that could serve as a second leg. All of these tracks were required to point to a fiducial volume of the muon detector, as defined by the MuonFiducialTool. If a track was part of a CdfMuon, then it passed, and if not, it failed. Ken then fits the J/psi peak in the passing and failing plots, and uses these numbers to calculate the efficiency. In the CMU and CMP, the efficiency was consistent with one, with an 8% uncertainty. People protested that if anything, this efficiency was way too high, given that there are chambers that are off and sometimes the fiducial tool will get it wrong. However, it could just be that the measurement is not precise enough to see this yet. In the CMX, Ken found a reconstruction efficiency of 0.231 +- 0.074. Now, the fiducial tool assumes that the arches and keystone are on, so this efficiency would probably only have a maximum value of about 0.75, but this is still bad. We need to understand what is going on with this, and to try to reduce the number of events in the "fail" category, to reduce the uncertainty on the measurement. Ken briefly described his attempts to measure the ID efficiency. This involved trying to reconstruct W's out of CdfMuons with very few cuts on them, and to estimate the background in this sample, and then to apply cuts one at a time and see how many events pass and fail (again, after a background subtraction). Some of the resulting numbers came out reasonable, and some did not; Ken is not yet confident of this methodology. ------------------------------------------------------------------------------ Ken Bloom, Karma Adjuster Department of Physics 734-763-2329 / 734-936-1817 (fax) University of Michigan kenbloom@umich.edu http://www.umich.edu/~kenbloom