280 likes | 422 Views
Measurement of the ϒ( nS ) μ + μ - Decay Angular Distribution Questions from pre-blessing (part 2). http://www-cdf.fnal.gov/~mjones/internal/Upsilon.html. Matthew Jones Purdue University. Question #5. Effect of FSR on signal: split 1S peak
E N D
Measurement of the ϒ(nS)μ+μ- Decay Angular DistributionQuestions from pre-blessing(part 2) http://www-cdf.fnal.gov/~mjones/internal/Upsilon.html Matthew Jones Purdue University B Production and Decay Meeting
Question #5 • Effect of FSR on signal: split 1S peak • Answer #5… the differences just looks statistical. • The width of the peak is dominated by resolution, not radiation • Look at other pT bins… B Production and Decay Meeting
Question #5 They all just look like statistical variations associated with splitting the sample in half. B Production and Decay Meeting
Question #6 • Effect of B background structure moving under the upsilon peak? • Answer #6… the structure in the toy Monte Carlo might not be described well by a polynomial: B Production and Decay Meeting
Question #6 • In fact, the distribution is quite non-uniform: Collins-Soperframe S-channel helicity frame B Production and Decay Meeting
Question #6 • When sliced into bins of |cosθ|, the distribution is more smooth, but a polynomial never fits very well. B Production and Decay Meeting
Question #6 • But whether it produces a bias is not so clear: • Fits to toy background + Gaussian with statistics similar to what is used in CDF data are not egregiously biased. • Some higher order polynomial fits don’t converge easily. • Only tried with one signal, not three • Difficult to check in data without a like-sign di-muon sample. • So can it lead to a bias? • I don’t know… • If it does, it is probably only in certain bins in pT and cosθ. B Production and Decay Meeting
Question #8 • Jonathan would like to see the analysis of the toys for the difference between the results in the two frames. • Answer #8… Here’s how the toy MC works: • In a given pT range, load the un-polarized CS and SH acceptance templates • Generate decay kinematics using polarization parameters in either CS or SH frame • Accept/reject event based on acceptance found in corresponding template • Simpler than using detector geometry and trigger parameterization • Fill histograms with both CS and SH angles for accepted events. • Do this for both signal and background including • Fraction of background in prompt/displaced samples • Fraction of signal in prompt/displaced samples • Fit distributions using same likelihood fit used in analysis of data • Additionally, randomize the templates using binomial pdf to simulate the effect of finite Monte Carlo statistics. B Production and Decay Meeting
Question #8 • Distributions of λfit-λtrue: Fit using identical acceptance used in generation: no contribution from Monte Carlo statistics. Fit using randomized acceptance: this includes contribution from Monte Carlo statistics. B Production and Decay Meeting
Question #8 • Using this procedure, we can account for how much of the uncertainty from the fit was from finite Monte Carlo statistics: • This is just a re-distribution of the uncertainty from the fit… it does not change the total uncertainty. B Production and Decay Meeting
Question #8 • But Jonathan asked about • Some bias is not surprising • Acceptance in one frame is exact by construction, the other is approximate • Width of distribution indicates how much variation can be expected purely from statistics This is for the 1S state. Toy MC jobs for the 2S and 3S states are still runningon namgrid… B Production and Decay Meeting
Question #8 • Is the variation statistical or systematic? • In many cases it must be statistical • In some cases it must be systematic • I think we should include it as a systematic in all cases. • When there is agreement it is negligible. • When there is disagreement, we need to quantify it. B Production and Decay Meeting
Question #10 • Sample correlation coefficients for talks • Answer #10… here they are: B Production and Decay Meeting
Question #10 • These actually aren’t really that interesting for a talk. • It would probably be better to show a triangle plot: B Production and Decay Meeting
Question #15 • Perform analysis in fewer pT bins to facilitate comparison with other measurements • Answer #15… Two alternate sets of bins considered: B Production and Decay Meeting
Question #15 14 bins 8 bins 5 bins B Production and Decay Meeting
Question #11 • Comparison with other results • Answer #15… Previous results: • CDF Run I publication: for the ϒ(1S) in the SH frame • DØ Run II publication: for the 1S and 2S in the SH frame • Blessed CDF Run II result: for the ϒ(1S) in the SH frame (CDF Note 9966) • Un-blessed CDF Run II results: CDF Note 10154 B Production and Decay Meeting
Question #11 • More or less consistent with Run I result • Not too surprising… the error bars were really large. B Production and Decay Meeting
Question #11 • Inconsistent with DØ result • Nothing new here… this is already a well known discrepancy B Production and Decay Meeting
Question #11 • Barely consistent with DØ result • Not too surprising given the previous ϒ(1S) result • This comparison has not been made in public before B Production and Decay Meeting
Question #11 • Inconsistent with previous blessed result. • This CDF II result has been widely circulated in public. • Not completely clear how to explain it away… B Production and Decay Meeting
Question #11 • This is in better agreement, but the result in note 10154 disagrees with the previously blessed result. • Not clear how to use this comparison since note 10154 is not blessed. B Production and Decay Meeting
Question #11 • First ever comparison that isn’t in the SH frame! • The level of agreement is not great… • If note 10154 result has a problem it should show up in B Production and Decay Meeting
Question #11 • If it isn’t the background, could it be the templates? • Selection criteria used in note 10154 is worrisome: Implicit requirement on silicon hits. Efficiency expected to depend on |cosθ| B Production and Decay Meeting
Question #11 • Is this really a problem? I just don’t know. • The Monte Carlo might get this right… • Reasons to suspect that it might not (hit resolution modeling, time-dependent silicon detector coverage) No σLxy cut With σLxy<0.25 mm Obvious structure from the silicon detector, asymmetric in Z. Efficiency does depend on cosθ B Production and Decay Meeting
Question #16 • What about rotational invariants other than ? • Answer #16… Both are referred to in the literature: F They really are fundamentally identical – no need to do both. B Production and Decay Meeting
Summary • I think all questions have been addressed • Main conclusion from the analysis: • No strong polarization even at high pT even for the 3S state. • The polarization was not “hiding” in the CS frame. • Maybe polarization evolves significantly with |y|? • Preparing final tables of numbers • 2S and 3S toy Monte Carlo jobs still running • Jim suggested putting error bars at mean pT rather than bin centers. This is a very good idea. • Can I please use MINOS errors? They are the same as parabolic errors in most cases, but in a few cases they are superior. • Sticking “CDF II Preliminary” on a bunch of plots… • Web page and complete set of plots/tables will be ready by Thursday • If it pleases the B group, we would like to proceed to blessing so that this can be shown at the QWG 2011 workshop, October 4-7. B Production and Decay Meeting