Skip to content

Neutral truth matching suspicious behavior

Inefficient truth matching in photons (even high PT signal photons)

The amount of photons being matched to ID 22 is presumably much lower than it should be, as seen in this plot of KstGamma MC where lots of events under the mass peak have the photon associated to different things.

Unclear whether it is a configuration problem that got me to that result or a more intrinsical problem on the matching side

image

Unstable truthmatch result of photons and large difference of mismatched photons fraction between different Moore(and related software) versions

I've run into 2 kinds of photon truth match problems when checking Hlt2RD_BdToKstGamma and Hlt2RD_BsToPhiGamma lines with an upgrade simulation file (sim10aU1). The tuples are produced by DaVinci v63r6 and Moore v54r2-v54r8p1

  1. The fraction of photons matched to \gamma_f in \gamma\to e\to\gamma_f processes is significantly higher when running reconstruction with Moore version after LHCb!3963 (merged) being merged (<10% before this MR (r_e=N_{true \gamma}/N_{\gamma e\gamma}=0/10, \~30% after this MR (r_e=14/24)). This problem also exists when checking with BuToJpsiKstPlus and BuToChicKPlus Htl2 lines and signal MC.
  2. The trueID distribution of photons is not stable in different Moore versions. For example, below are the trueID distribution of photons from same xdigi file, triggered by the same HLT2 line, reconstructed with Moore v54r6 and v54r8p1 separately. Moore v54r6: image Moore v54r8p1: image I also scanned all entries on these two plots, found that the photon information from the reconstruction side (the CaloID, reconstructed energy, Calo energy, run number, and event number) is the same when running with different Moore versions while truth information is different. Moore released versions from v54r2 to v54r8p1 are checked, and the truth information keeps changing in different versions.

The xdigi file I used to do the version consistency check is /lhcb/MC/Upgrade/XDIGI/00153942/0000/00153942_00000154_1.xdigi. The related xdigi->dst and dst->tuple scripts can be found at /afs/cern.ch/user/y/yihou/public/neutral_truthmatch, as well as the tuples and entry scan with different Moore versions.

FYI, @graven @jzhuo @chefdevi @jmarchan @yihou @ibachill @bdey

Edited by Aniol Lobo Salvia