Project

General

Profile

What we know about the E-per-hit discrepancy

Statement of the problem

Since olden times, we've had a data-MC discrepancy in the mean calorimetric energy per hit in reconstructed ND nue events where E/hit is predicted to have a harder spectrum by the MC than what we observe in data. In general, it seems that the origin of this discrepancy is that the simulation is producing too few hits; the total energy of selected events is modeled well in every case except for the Second Analysis with LEM. See:

  • First analysis: Doc 13587, figs. 33-34 (LEM selection) and figs. 50-51 (LID selection)
  • Second analysis: Doc 15396, fig. 15 (LEM selection), fig. 29 (LID selection), fig. 37 (CVNe selection)
  • 2017 analysis: Doc 22475, figs. 5d and 8a
  • 2018 analysis: Doc 26664, fig. 2

However, the discrepancy for selected candidate showers themselves has evolved over time. This is possibly in part because of changes in the xsec central value tune.

  • In the first analysis, shower energies were typically underpredicted in MC (Doc 13587 fig. 35, LEM; fig. 52, LID)
  • In the second analysis, shower energy prediction depended on classifier; Nhits not reported (Doc 15396 fig. 18, LEM; fig. 29, LID; fig. 40, CVNe)
  • 2017 analysis: shower energies predicted well, Nhits underpredicted in MC (Doc 22475, fig. 4)
  • 2018 analysis: shower energies predicted well, Nhits underpredicted in MC (Doc 26664, fig. 3)

Dedicated studies

  • Emrah Tiras developed an art module to study hit-level variables of selected EM showers: summary Doc 31345.

In other EM showers

  • MRBrem
    • 2017 summary in Doc 23165. Shower energy, length modeled very well. Shower width perhaps underpredicted by MC? NHits slightly underpredicted as well.
    • 2018 summary in Doc 30222. Fewer variables shown than previous.
  • Pi0 studies
    • Micah G., Doc 27314 found that the pi0 sample used for calibration xchecks exhibits discrepancies in both shower energy and nhit distributions. Here, however, both energy and nhits are higher in MC than data.
    • Dan P.'s ccpi0 cross section analysis, Doc 18768 uses E/hit as a part of its primary PID. Early studies, e.g. Doc 16163, p. 5, show reasonably good agreement in this variable, with MC only slightly overpredicting, in a fairly pure sample. However, this was SA MC, so it is unknown how well these conclusions extrapolate to the differently-tuned light levels of prod3+4.