Study of Low Y Data/MC Disagreement

Introduction

It was noted at the Oxford collaboration meeting (Jan 06) that there was disagreement between R1_18_2 processed data and carrot MC in the ND at low y (and low reconstructed neutrino energies). Figures 1 and 2 are a reminder of the energy spectra and y distributions for data and MC before and after a QE-like selection procedure has been applied. The QE-like selection locks onto the low y region and the level of disagreement in this sample is a big problem for the ND flux shape measurement. The following cuts have been applied to get these samples:

• MC: (6.71e18 POTs)
  • reco_enu < 20 GeV
  • pass==1 (one good quality track and Pitt fiducial volume cut)
  • trkqp <= 0
  • if muon momentum is measured from curvature (trkeqp/trkqp) < 0.3
  • reco_trk_vtxz > 0.5 m
  • no more than 20% of an event's PH or hits are re-used in another event in the same snarl
• Data: (9.75e19 POTs)
  • all above cuts
  • standard beam monitoring cuts
  • special horn current runs excluded

Figures 1. Energy Spectra

• The scales for the energy spectra ratio plots are completely wrong (something to do with the pre-scaling of the histos) but the shape is accurate.

Figures 2. y Distributions

Initial Hand Scan

• My intial effort to understand this problem was a hand scan of data and MC events with reco_y < 0.2.
• I scanned ~100 events in each sample to get a feel for how the reconstruction was handling the data and MC and to see if there was any gross and immediate problem.
• I felt that the reconstruction was doing well in both cases and that there was no readily apparent difference in the way that data and MC were treated.
• This initial hand scan was dominated by higher energy events where the data and MC agreement is good.

I made the following observations whilst scanning the events:

• There are very few events with zero hadronic energy as there is often at least a one or two hit Brems along the muon track. As such a zero hadronic energy selection would need to check the shower vertex against the track vertex before using the largest shower in an event for the shower energy.
• There are many events where there is no visible vertex hadronic shower but a large (~500 MeV) Brems along the muon track - does this get used to calculate y?
• It felt to me like there may have been more cross-talk like hits in the MC than in the data but the higher reconstructed neutrino energy events could have been fooling me.

I did not see any 'smoking guns' and so carried out another hand scan on samples of low y data and MC events with reconstructed neutrino energies less than 4 GeV...

Second Hand Scan

• In this scan I considered 50 events from both the data and MC samples.
• The reconstruction again seemed to be doing well in both data and MC with no obvious differences in the was they were being treated.
• One thing that I wanted to check in this hand scan was whether or not there really were more cross-talk like hits in MC than in data. I counted by eye the numbers of hits in the vertex hadronic shower of each event and also the numbers of these hits with PH < 2pe.

The following figure shows the vertex hadronic shower hit multiplicities from the hand scanning for data and MC (normalised to number of events included - there are only 46 events in MC as I ignored some junk events that came in the hand scan sample):

Figures 3. Hadronic Shower Hit Multiplicities

The means of these histograms do not look particularly different given the poor statistics. I decided to re-make pans for the data and MC including variables that counting hits in the hadronic shower to provide a more quantitative result on the levels of cross-talk and hit multiplicities in data and MC. These results follow in the next section.

Again I made some observations (and formulated some questions) during the hand scanning:

• For these event there is little or no hadronic energy and I am still worried that the large (~500 MeV) Brems seen in many of the events are being used in the y calculation. If this is true could there be some difference in the number/size of Brems showers between the data and MC?
• Do shower hits that also constitute part of a track have any energy removed before they are included in the shower energy calculation?
• Is shower completeness and purity used at any point?

Trying to Quantify the Observations

The following plots will compare the R1_18_2 processed standard LE-10 MC and data from June. For each variable I have plotted data and MC both before (1st plot) and after (2nd plot) my QE selection procedure and with both POT and event normalisations. I have used the same selections as described in the introduction but with:

• reco_enu < 4 GeV
• reco_y < 0.2

For reference the plots of reconstructed neutrino energy and y for this selection look as follows:

Figures 4. Energy Spectra

Figures 4a. y Distributions

Firstly I wanted to look at some event hit multiplicities. Figures 5 show the hit multiplicities for hits that I classify as part of the vertex hadronic shower, this includes hits that are:

• Either part of a shower or part of a shower and track.
• Have pulse heights above 1.5 pes (an attempt to remove cross-talk like hits).
• Are not more than 2m in z from the track vertex (protons and pions should not travel further than this).

Figures 5. Vertex Hadronic Shower Hit Multiplicities

And the corresponding PH totals (in SigCors) for these hits:

Figures 6. Vertex Hadronic Shower Summed PH

Figures 5 can be contrasted with the following figures 7 that show hit multiplicities for any hits in an event that are:

• Part of a shower.

Figures 7. Shower Hit Multiplicities

The next set of plots deal with shower multiplicities and energies. Figures 8 show the numbers of shower in events and figures 9 the total number of sub-showers (of all classifications):

Figures 8. Shower Multiplicities

Figures 9. Sub-Shower Multiplicities

Figures 10 show the shower energy spectra and figures 11 the muon shower energy spectra:

Figures 10. Shower Energies

Figures 11. Muon Shower Energies

Finally I wanted to look at the difference between the track and largest shower vertex in an event:

Figures 12. Shower Vertices

And in log form:

Figures 13. Shower Vertices - Log Scale