Run 9 200GeV Dijet Max Jet Pt Cut Investigation

Look at data / simu jet pt comparisons to determine upper jet pt cuts ...

The comparisons shown use runs from the RFF and FF portions of the run.

Figure 1: This figure shows the data (Blue) and simu (Red) jet pt spectra and data/simu ratio for the 6 different trigger categories.


I originally made these comparisons to nail down where the cuts on jet pt should be placed, but I noticed that the data/simu ratio for the JP1Hi and UnJP1Lo triggers deviated more than in previous comparisons. I thought that the deviations may be due to the inclusion of production2009_200Gev_Hi and Lo triggers from the early part of the RFF run.

Figure 2: Same as figure 1, but now I have removed the Hi and Lo runs from the data sample so only the production2009_200Gev_Single remains. 


Figure 3: Same as figure 2, but now I have removed the Hi and Lo runs from the simulation sample as well.


The Hi and Lo triggered runs make up a non-negligible fraction of the runs I analyze, here are the number of dijet in the full and no Hi or Lo samples:

  • Data All = 12218840, Data No Hi or Lo = 11144220 (8.8% loss)
  • Simu All = 290689, Simu No Hi or Lo = 283228 (2.6% loss)

I have also looked at the same-side and away-side jet pt spectra seperately to see where I should place the jet pt max cuts. The figures below are for the data sample with Hi and Lo triggered runs removed and the full simulation sample.

Figure 4: This figure shows the data/simu comparisons for the same-side jet pt. Note that untriggered jets cannot be same-side by definition.


Figure 5: Same as figure 4, but now for away-side jets.


I will cut events with jets satisfying the following conditions for all subsequent analyses:

  • JP2        > 45 GeV
  • JP1Hi     > 40 GeV
  • UnJP2    > 45 GeV
  • UnJP1Lo > 30 GeV
  • UnJP1Hi > 30 GeV

============================================
============================================
============================================

Continuing my investigation into the data / simu mismatch in JP1Hi and why removing Hi and Lo trigger runs seems to fix the issue ... 

The full runlist I use in this analysis can be found here. It contains 802 runs. The first 95 runs (up to and including run 10125027) consist of production2009_200Gev_Hi and production2009_200Gev_Lo trigger setups. Of these 95 runs, 20 are Lo and 75 are Hi

There was a question as to weather the mismatch seen could be due to an Endcap problem and there is indeed 1 jet patch which is low for the entire duration of the Hi | Lo run period.

Figure 6: This figure shows the Jet Patch plot from the PPlots from run 10120032.

To see if this low jet patch could be the cause of the mismatch, I made a data / simu comparison where I look only at jets with a detEta < 0.9 and no Endcap towers included.

Figure 7: This figure shows the jet pt data / simu comparisons for the 6 trigger categories for jets with detEta < 0.9 and no endcap towers.


The mismatch can still be seen even though the endcap was not included. I was not able to exclude event triggered by the endcap without rerunning my trees, so there may still be an effect from the bad jet patch but I think it is unlikely.

My original hypothesis was that the mismatch was caused by some sort of prescale issue. I looked at the Lo trigger setup runs and in each one, the JP1 (which goes into JP1Hi) trigger was prescaled. I didn't look at all the Hi trigger setup runs, but I don't think L2JetHigh was prescaled or if it was it was rare. To see if the mismatch was due to JP1 being prescaled in the Lo configuration, I made a data / simu comparison where I leave out the 20 production2009_200Gev_Lo runs but include the 75 production2009_200Gev_Hi runs.

Figure 8: This figure shows the jet pt data / simu comparison for the 6 trigger categories with the production2009_200Gev_Lo runs excluded.


As seen above, removing the 20 Lo runs fixes the mismatch about as well as removing all 95 Hi and Lo runs as in Fig 3. If the issue really was with the endcap jet patch, I would expect the mismatch to persist in figure 8 because there are still 75 runs with the endcap problem included. This leads me to believe that the mismatch was caused by the prescaled JP1 trigger in the production2009_200Gev_Lo runs. Here are the stats for the removed Lo runs:

  • Data All = 12218840, Data No Lo = 12061040 (1.3% loss)
  • Simu All = 299249, Simu No Lo = 295964 (1.1% loss)