Updates Since Sept 13th

General updates since my Spin PWG talk of Sept. 13: additional systematics, trigger bug fix, additional cut, various adjustments, and studies regarding the trigger.


Overview of Updates:

  • Updated to newest polarization values
  • Templates fit in wider mass range
  • Decision regarding pT range for final plots (for DNP)
    • Asymmetries: pT 4-16 GeV
      • Previously used Weihong's choice of 4-12 GeV, but expect run 9 to have better contraint on 12-16 GeV bin, and would like a run6 value with which to compare it.
      • Both are shown below: 4-12 GeV for comparison with Weihong, and 4-16 GeV for proposed "money plot"
    • Cross section: unfold in 5-16 GeV, but only plot 6-16 GeV
      • Hope to get one more pT bin for the paper after solve trigger simulation questions
  • Checked the background asymmetries in two mass regions.  See more details here.
  • LW Points for the cross section horizontal position of points
    • Fit all points to exponential using combined stat. & sys. uncertainty, and then compute LW points based on the fit
    • Reduced chi^2 is less than one
    • Reference: Nuclear Instruments and Methods in Physics Research A 355 (1995) 541-547.
  • Additional Uncertainties
    • Both cross section and asymetries
      • Variation due to fit range (nominal is 0.1-0.5 GeV, took variance of moments of 6 combinations: low edge at 0.1 or 0.11 GeV and upper edge at 0.45, 0.5, 0.55 GeV)
    • Just cross section (more details below)
      • average pT
      • 2% energy scale uncertainty
    • Scale uncertainties
      • Asymetries (due to polarization) taken from here.
      • Cross section: 7.7% scale uncertainty = 2/26.1 (bbc sigma over cross section)
      • Both are now included on the plots
  • Found bug in reading results of trigger simulator--before only using L0 portion, now use both L0 and L2.
  • Running on newest runlist from Hal, restricted to last trigger threshold setting and runs with the TPC
  • New mock up L2 cut (treated as a reconstruction cut).  See more details here.
  • Checked relation between Delta R < 0.04 and RMS of Delta R distribution (used when associating data & MC pi0s).  More details here.
  • Updated trigger efficiency based on newest cuts & trigger fixes

Remaining Items to Address:

  • Add theory curves (Marco Stratmann has promised cross section curve by Sept 22--updated to Sept 29)
  • Check effect of weighting MC based on vertex Z distribution (in progress)

As an aside, it was also suggested to check the eta dependence of the alpha energy scale factor.  This study is not directly related to the results, but does relate to the discussion regarding whether the EEMC fast/slow simulators are using the best choice of sampling fraction.


Comments Regarding New Cross Section Systematics

Average pT

The absolute resolution on the EEMC energy has been calculated to be 0.16/sqrt(E).  Each bin of the cross section has a 1/<pT> factor.  The uncertainty on <pT> is set to 0.16/sqrt( cosh(1.1) pT ) per bin.  The nominal value of 1.1 is used as, 1) the resolution is worse at lower pT, and 2) using the lowest eta value would over estimate the uncertainy.

EEMC Energy Scale

The newest fits result in a peak position of 134.7 MeV (PDG has 134.98) with an RMS of 3 MeV.  This is a 3/134.7 = 2.2% relative uncertainty.  The uncertainy on the alpha parameters increase with pT, due to the drop in statistics.  The maximum uncertainty is in pT 12-16, with value 0.013.  The next pT bin (10-12 GeV) has uncertainty 0.007.  The uncertainy on alpha suggests a higher energy resolution than suggested by the variation of the peak position.  Based on these values, we decided to quote a 2% energy scale uncertainty.

To determine the associated systematic for a given pT bin, we

  1. Assumed the cross section is locally exponential
  2. Fit two bins of the cross section to an exponential.  This is an NDF=0 fit--two parameters, two data points.
    1. The bin under consideration and the next lowest pT bin, if there is one
    2. Otherwise, the bin under consideration and the next highest pT bin.
  3. Integrated the exponential between +/- 2% of the upper bin edge (the edge between the two points of the fit)
  4. Assigned half this value as the systematic uncertainty

We only considered the upper bin edge, as this dominates over the uncertainty which would have been computed at the lower bin edge.


Study Regarding Effect of Data/MC Mismatch

The entire analysis was processes using the MC with the mockup L2 trigger set at 3.7 and 5.2 GeV (used for the following final results) in the attachements, and also at 1/0.97 times the thresholds used for the data (2 and 4 GeV).  Attached at the bottom of this blog are the plots for the data/Monte Carlo comparison with both these settings: 3.7 & 5.2 GeV is denoted "f24-f0" in the file names while 2/0.97 & 4/0.97 is denoted "f24-f24b" in the file names.  Note: the comparison on this blog did not have the correct relative normalizations for the vertex Z distribution, due to the bin width being different.  This is now corrected.

The difference in the final results for the asymetries is a full order of magnitude less than the statistical uncertainties, and is thus negligible.  For the cross section, the difference is statistically significant, but is negligible compared to the present systematic uncertainties.  For example, adding a systematic equal to half the difference in the final results would increase the systematic error in the pT 6-7 bin by 2%.  Higher pT bins would be effected even less.  While for the publication, we will work to improve the data/Monte Carlo comparison and try to solve the current puzzle (why the mockup L2 sees events one would have expected the simulated L2 to have cut), the systematics seem to accurately account for the uncertainty within the chosen pT range of the plot.

Based on this study, it is expected that reweighting the vertex distribution will also have negligible effect on the final results.


Updated Cross Section


Comments Regarding Polarization Values

I've checked the polarization values between that which I am using and that which is listed in Appendix B of Weihong's Thesis (the version posted at STAR).  All checked values are in agreement, so I believe we are now using the same polarization values.  However,  a line in the text (p. 130, or p. 145 of the pdf) states his polarizations are 54% (blue) 56% (yellow).  This is not consistent with either the luminoscity weighed average or the direct average of the values in Appendix B.The value quoted by Scott in a recent email (54%) agrees with blue value of the polarization, but not the yellow.  It is not known whether the values in the appendix or text match the values in the code.  Thus the issue raised by Hal in this email (and said to be closed by me in this email) is still open.  While it is probably not worth pursuing further, it should be noted:

  • There is question of internal inconsistency in Weihong's analysis.
  • The single spin asymetries in Weihong's theisis and shown by Scott at the Spin 2008 are the RAW asymetries, not the A_L values.
  • The lumi-weighted average polarization I am using is 57% (blue), 59% (yellow) and 33% (product), which are consistent with Appendix B.

Updated Asymetries, pT in  4-12 GeV (to compare with Weihong)

 

Note: fitting A_LL to a constant yields 0.040 +/- 0.020 with a chi^2/ndf = 5.8/6.


Updated Asymetries, pT in  4-16 GeV ("Money Plots")

Note: fitting A_LL to a constant yields 0.039 +/- 0.020 with a chi^2/ndf of 6.1/7.