From stark@SLAC.Stanford.EDU Mon Nov 10 10:54:39 2003 Date: Thu, 16 Oct 2003 01:39:26 -0700 (PDT) From: Jan Stark To: d0calgo@fnal.gov Subject: Low-pT di-EM sample Hello, As you know, trigger list version 12 includes a new trigger for low-pT di-EM events. One of the signals that this trigger can pick up is J/Psi -> e+ e-. At the "All D0 meeting" (ADM) on August 15th I have presented a quick look at the first data that have been recorded with this new trigger (7.6 pb-1). The slides of this presentation are on the web: http://www-d0.fnal.gov/atwork/adm/d0_private/default.htm The purpose of this mail is to give an update with the full sample that we have integrated before the shutdown (38.4 pb^-1), and, more importantly, to provide you with a convenient way to access this sample and use it for your detector studies. The description below should allow you to use this sample. Please let me know if something is not clear or if you run into trouble. On October 28th we will have a special Calgo meeting devoted to EM resolution and simulation. This sample may be a useful tools for studies related to these two topics. Cheers, Jan --------------------------------------------------------------------------- The trigger is described in the ADM talk and references therein. The integrated luminosity corresponds to 38.4 pb^-1, reconstructed with p14 and the certified EM postprocessing (well, certified for p13). J/Psis can be reconstructed using different techniques to reconstruct the electrons, including the road method or the usual clustering method. The differences between these two methods are discussed in the ADM talk. Let's start with the road method. We require one loose road electron and one tight road electron. This is essentially the same requiment as the level 3 requirement of the trigger. We obtain this mass distribution: http://www-clued0.fnal.gov/~stark/jpsi/road.eps The mass of each J/Psi candidate is calculated from the momenta of the tracks that correspond to the road electrons. The number of J/Psi mesons in this sample is of the order of 2200. Superimposed on the data points is the result of an unbinned maximum likelihood fit. The signal is described by a "Crystal Ball" function (it models the radiative tail). The background is described by the sum of two exponentials. Monte Carlo results are discussed in the ADM talk. Let's move to cluster-based reconstruction. We require two EM clusters with: EM fraction > 0.9 (standard certified cut) isolation < 0.15 (standard certified cut) HMx8 < some value which we will vary In addition, we require each cluster to be within 0.2 in eta/phi space of a loose road electron (this is essentially a track-match). Let's start with HMx8 < 75 (very loose cut). This is the mass distribution we get: http://www-clued0.fnal.gov/~stark/jpsi/psi75.eps The number of J/Psi mesons in this sample is of the order of 730. As discussed in the ADM talk, this number is significantly smaller than the yield with the road method. With the clustering method and the isolation cut, we restrict ourselves to isolated J/Psi mesons, predominantly from direct production rather than from b decays. Superimposed on the data points is the result of an unbinned maximum likelihood fit. The signal is described by a gaussian. The background by a threshold function: f(m) = ( 1 - exp( - ( m - 1.1 GeV) / c ) ) where c is free to vary in the fit. Monte Carlo results can be found in the ADM talk, including a discussion of the trigger bias. We now tighten up the HMx8 cut. We replace HMx8 < 75 with HMx8 < 50 and get this mass distribution: http://www-clued0.fnal.gov/~stark/jpsi/psi50.eps We further tighten up the cut. We replace HMx8 < 75 with HMx8 < 20 (certified "tight" cut) and get this mass distribution: http://www-clued0.fnal.gov/~stark/jpsi/psi20.eps With this tight cut, we get 350 J/Psis with impressive purity. Let's stick with the tight HMx8 < 20 cut and enlarge the mass window. This is what we get: http://www-clued0.fnal.gov/~stark/jpsi/psiups20.eps We see the Upsilon as well. Unfortunately, neither our tracking system nor our calorimeter has a resolution that is good enough to separate the different members of the Upsilon family. This was neglected in the ADM talk. Here we do a slightly better job. The same problem is obviously an issue in the recent D0 measurent of the Upsilon(1S) production cross section in the Upsilon(1S) -> mu+ mu- channel, documented in D0 note 4264 by Bauer, Huang and Zieminski. For the mass fit shown in the previous plot, we use am Upsilon model that is inspired by the Upsilon(1S) -> mu+ mu- analysis. We have two components, Upsilon(1S) shown in blue and Upsilon(2S) shown in pink. Given the low statistics, we neglect the Upsilon(3S). The mass of the Upsilon(1S) is a free parameter in the fit, the mas of the Upsilon(2S) is calculated from the (floating) mass of the Upsilon(1S) and the value of the mass difference taken from the PDG. Both Upsilons are modelled as gaussians. The width of the Upsilon(1S) is a free parameter, the width of the Upsilon(2S) is calculated from the (floating) width of the Upsilon(1S) and a small correction factor to account for the expected resolution difference due to the mass difference. The Upsilon(1S) event yield is free to vary in the fit; the Upsilon(2S) yield is assumed to be 45 % of the (floating) Upsilon(1S) yield. The choice of 45 % is inspired by the Upsilon -> mu+ mu- analysis. Letting this fraction of events yields float gives a consitent result with a large statistical error. We now take exactly the same di-EM candidates and calculate the di-EM mass from the track momenta rather than the EMparticle momenta. The corrsponding plot is shown here: http://www-clued0.fnal.gov/~stark/jpsi/psiupstrk20.eps The functional form for the fit is the same as in the previous case, except for the model of the Psi. We use a Crystall ball function (radiative tail) plus an additional, similar component to accomodate for something that looks like a Psi(2S) signal (given the resolution difference, such a signal would be tough to see in the previous case). Some comments on this CAL - tracker comparison, and also on the comparison of the track case with the muon analysis, are in order: - As in the ADM talk, we see that the tracker and the calo do not agree on the Upsilon mass. But given that we now include the Upsilon(2S) contamination, the calo now looks better and the tracker result is now more suspicious (compared to the ADM talk). It is difficult to investigate this further with out any Upsilon -> e e Monte Carlo. This MC is currently in production. - The muon analysis (it uses muons matched to central tracks) sees a _much_ better resolution than we seen in the track case. Again, we need MC. - The background model used in these fits is too simple; it doesn't fit well over this large mass window. The background shape in the mass region of the Upsilon has some interesting features (which become more evident as one loosens the HMx8 cut). Need to do a better job on this, but would like to see MC first. The bottom line is that the J/Psi sample should be fairly simple to use; the Upsilon is much more tricky. The events selected using EM clusters, the loose HMx8 cut (HMx8 < 75) and without any mass cut can be found in this small and convenient thumbnail on clued0: /work/ladibona-clued0/stark/jpsi/jpsiups.tmb I am currently picking the RAW data for these events and will ask the CPS guys to reprocess them with a release that spits out decent CPS information. The picking is tape drive limited and my initial estimate is that this will take a while. Right now, there is no small thumbnail with the larger road-based sample, but I could make such a file quickly if you are interested. In the directory /work/ladibona-clued0/stark/jpsi/direct you can find a bunch of DSTs with Monte Carlo events: J/Psis from direct production (generated with pT(hard scatter) > 2 GeV and subsequent selection in d0mess). If you use _all_ of these files, you get a realistic (well, to zeroth order) mixture of the different direct production processes. I have the corresponding TMBs on a tape in France and can transfer them to clued0 if you are interested. The trigsim root-tuples (trigger list v12) for these events are in the directories: /work/ladibona-clued0/stark/jpsi/trigsim/TrigSimAna_x-p15.03.00-Linux-anaD1 /work/ladibona-clued0/stark/jpsi/trigsim/TrigSimAna_x-p15.03.00-Linux-anaD2 /work/ladibona-clued0/stark/jpsi/trigsim/TrigSimAna_x-p15.03.00-Linux-anaD3 /work/ladibona-clued0/stark/jpsi/trigsim/TrigSimAna_x-p15.03.00-Linux-anaD4 /work/ladibona-clued0/stark/jpsi/trigsim/TrigSimAna_x-p15.03.00-Linux-anaD5 /work/ladibona-clued0/stark/jpsi/trigsim/TrigSimAna_x-p15.03.00-Linux-anaD6 (there are fewer root files than DSTs, but the root files contain all the events from the DSTs). -----------------------------------------------------------------------------