1. Reconstruction Software Report J. Musser FNAL, Oct. 2005

2. Summary Progress Since Ely –AltdeMux adopted and used in R1.18 –SubShower adopted and used in R1.18 –Andy Culling’s shower energy calibrations used in R1.18 Near Term Plans –Address issues identified in R1.18 by analysis groups –Adopt J. Marshall’s Kalman Implementation –Consider adoption of CandChop Further Future Plans/Possibilities

3. 6 month Reconstruction Software Roadmap (From Ely) The next frozen release will include the following –Update/fixed shower energy scale. Done –Rough coil description in swimmer. Done –CandShield. Done (latest T0’s need to be added to dB) Over the next 6 months a number of SR packages will be replaced. Each of these will require the generation of evaluation data sets from the batch processing group, and validation by the reconstruction and analysis groups. –AltDeMux Commission data set immediately, target Aug 1 for adoption. Done –CandSubShower Commission data set ~ July 1, target Aug 15 for adoption. Done –CandChop Determine whether far detector spill trigger processing is degraded due to demuxing issues. If so, consider immediate adoption. If not, consideration of this can wait until other new code is in place. –FittersPush on getting the final swimmer in place. I need to work with Sue to understand final issues. Determine means for providing the results from multiple fitters (SR, SA), to facilitate studies of systematics, and resolve these. There is a date at which the software must be frozen for use in the ‘fall’ study. It’s important that we have as much of the new stuff in at that point, and that it be solid. Support of the energy scan will include MC studies to first understand the reco performance in the two unstudied configurations, with follow-on work to tune/improve all aspects of the reco code. This will involve most if not all of the ‘core’ reco group members.

4. ND Reconstruction Slicing in Beam Scan (Tom) Tom O. - Shows # slices generated vs POT in three beams. No evidence of non-linear behavior.

5. ND Reconstruction Slicing in Beam Scan (Tom) Total Sliced PH vs POT. Non-linearity observed in pHE is thought to be largely artifact of profile distribution.

6. ND Reconstruction Slicing in Beam Scan (Tom) Distribution of event pulse heights is independent of Intensity in all beam configurations. Conclusion: No Evidence of rate/beam-dependent Slicing efficiencies over range of intensities probed.

7. ND Reconstruction Energy Scan – Tracking (Niki)

8. Energy Scan - Summary No reconstruction issues were identified that could explain large MC/data energy shift seen in pME and pHE beam data. (known now to arise larger in MC generation) Some degradation in reconstruction IS seen as one goes to higher energy beam configurations. Whether this could be responsible for some of the residual MC/data offset seen is an open question.

9. Shower Reconstruction Update

10. Hadronic Energy Scale CC events, with showers calibrated using linear calibration. No obvious departures from Reco=True trend line. True Shower Energy Reco Shower Energy (Far Detector)

11. Hadronic Energy Scale True Shower Energy dE/E CC Events, Far Detector. At low energies, response function becomes quite non-gaussian. The parameterization used to convert shower sum to energy places (Ereco-Etrue)[most likely]=0 This causes (Ereco-Etrue)[ave]!=0.

12. Hadronic Energy Scale (High Energy) dE/E FD CC events Ereco-Etrue/Etrue Shower Energy > 5 GeV

13. Hadronic Energy Scale Shower Energy dE/Sqrt(E) Niki has reported an offset in E(reco) relative to E(true), at levels of ~ 25% @ 10 GeV in the ND.

14. Hadronic Energy Scale In tracking this down, one should keep in mind the following… –The shower calibration itself is identical for the far and near detector data. Therefore, a Near/Far difference as seen here requires –Near/Far Shower completeness differences at the level of 25% @ 10 GeV.

15. Shower Completeness vs Evis Far Near 100% 70% 60% 100% Visible Energy(GeV) 0 2 4 6 8 10 12 14 Visible Energy(GeV) Black: SS Red: SR

16. Hadronic Energy Scale A possible explanation In the Near Detector, matching the truth info with the reconstructed event is significantly more complicated than in the Far detector. An error In this association will lead to behavior seen by Niki. At right is a plot of dE/E in the Near Detector for Eshower>5 GeV, with special care taken to make sure that association is correct. No offset I seen. dE/E (E>5 GeV)

17. Event Time & LOW PH PEAK IN 1.18 Event time has grown bigger because of SS, in the sense that already the slicing was maybe too loose, but hits were not previously peaked by track or shower (I will re-visit Slicing in the near future) My previous Event timing cut ( and various combination of timing cuts) will not eliminate now the vast majority of Low PH excess events (which means “healthy” and problematic activity is now more mixed up together). Event Time Width (Data Black, MC: Red)

18. SS Cluster Info & Low PH Peak The strongest point (from what I have seen) of SS shower finding is the Cluster Information. I used this information to impose cuts on showers and the cleaning of the DATA is quite successful. Place a cut on percentage of Shower made up of “UNKNOWN” “CROSSTALK” OR “HALO” clusters. Total PH NO CUT Total PH MC DATA SS Cluster ID cut MC DATA

19. SS Cluster Info & Low PH Peak Con’t Agreement between data and MC much better … However the basic problem remains : This excess activity is constantly present affecting also “healthy” events and it would be nice if it could be safely eliminated… # Showers# Tracks Event Length PH per Plane Shower PH per digit Shower PH per digit lower in data but HUGE un- physical peak at ~ 100 ADC counts removed

20. Multiple Strip Removal : Method & improvements Elizabeth Barnes is an undergraduate (starting her 3 rd year) student at the University of Minnesota and has been a Fermilab – MINOS summer student for 2 summers. She managed to reproduce the excess of low PH events and the fact that they mostly constitute of strips in the outer PMT pixels by looking at strips that appear more than once in reconstructed neutrino events. Elizabeth's talk describing in detail what she did can be found in: 09 08 05 Reconstruction Phone Mtg http://home.fnal.gov/~niki/ebarnes/reco_09_08_05.pdf I “coded” that in CandStripSR (namely I don’t consider digits in the strips formation process that have already appeared in a previous Cand Strip) and examined the results in both data and MC.

21. Multiple Strip Activity Data & MC : Some Numbers Frequency of multiple Strips DATA (LE-10) : 10 % MC (LE-10) : 4% Frequency of multiple Strips DATA (LE-10) : 45% MC (LE-10) : 6 % in events with Total PH <10000 Frequency of multiple Strips DATA (LE-10) : 6% MC (LE-10) : 4% In events with Total PH > 10000

22. Multiple Strip Activity Data & MC : Timing The vast majority of the excess strips in the data has a “decay” time of 2 usec. In order to avoid removing healthy strips and introduce energy biases I modified my initial code to remove multiple strips if occurring within 2 usec from initial strip. Time difference (sec) of n-strip from first strip for events with PH < 10000 ADC counts Black DATA, Red MC (normalized to POTs)

23. Number of Tracks/Showers Multiple Strip Cut Applied # Tracks/Event # Showers/Event

24. Shower Pulse Height Shower Pulse Height (MC) (Data) MC shower pulse height distribution uneffected by cut, while in data low pulse height events strongly suppressed.

25. Neutrino Energy Spectrum and Y Distribution Effect of cleaning cut »LE Energy »Y Multiple Strip cleaning cut has no effect on MC E, Y dist.

26. Event Vertex in Rock Muon Events In some cases and in particular in many rock muon events, the event vertex is being mis-placed Shower/Event Vertex Track Vertex

27. Shower Reco Conclusion Decision should be made to implement low pulse height removal cuts that is agreeable to all analysis groups. Sub-Shower code has different time requirements than old SR that has introduced some problems with increased backgrounds in FD spill event sample. This will be fixed. (Chris) Event vertex being mis-placed in some rock muon events with brem. This will be fixed. Problems with ND shower energy reco reported yesterday probably understood, but should be completely nailed down.

28. Tracking Update

29. Muon dE/E PLE dE/E LE Muon Energy dE/E vs E Some dependence of dE/E on E is seen, with offsets of as large as 10% at very low energy.

30. B-Field Scan (SR and SA)

31. New Kalman Filter Implementation J. Marshall’s fitter is close to ready for a validation procedure leading to adoption as the standard fitter. The new fitter appears to reproduce SR performance characteristics, with many improvements, including… –Improved track finding –Lower failure rate –Faster

32. Extra Track Finding Simply matching the strips to the Kalman state vectors is not enough when we must deal with very poor track finding in a vertex shower. In these cases, we count the number of hits on a plane, in order to identify the plane that divides the ‘clean’ section of track from the track hits inside the vertex shower. We trust the clean section of the finder track and we use it to allow the Kalman filter to build up information about the muon. This allows a more informed and a more successful track strip identification on the planes within the vertex shower. ‘Clean’ section of track Find track hits within shower using information from clean part of track Strips found by CandTrackSR Need to identify this plane

33. Extra Track Finding – Cont’d When we reach the plane separating the clean track and the shower, we use the Swimmer to predict the most likely track strips, adding hits one plane at a time to the Kalman fit. After this identification, we can simply move forwards and backwards along the track, as before, and obtain the final output. Blue: Strips from the track finder, mostly covered by… Red: final fitted track strips True muon hitsReconstruction

34. FD Beam MC Fitting failures: New: 0.2%SR: 3.0% Red: New Fitter Points: Latest SR Red: New Fitter Mean: 0.1559s Black: Latest SR Mean: 1.1960s All SR q/p values are ‘tweaked’ to remove q/p offset. Have done this to new fitter results too, for sake of comparison. 20,000 Snarls

35. ND Beam MC Results seem excellent, and the new fitter is no slower for ND than for FD. Red: New Fitter Points: Latest SR Red: New Fitter Mean: 0.1564s Black: Latest SR Mean: 1.341s Fitting failures: New: 2.4%SR: 9.1% 2,200 Snarls

36. FD Cosmic MC Note different number of entries in range {–1,1}. Due to resolution values lying outside this range, together with larger number of fitting failures. Red: New Fitter Points: Latest SR Red: New Fitter Mean: 0.1160s Black: Latest SR Mean: 0.7550s SR error estimation clearly good, as outlying points brought back into range by large QPVtxError. Fitting failures: New: 0.1%SR: 1.5% 100,000 Snarls

37. Swimmer Mods (Armstrong) Bob Armstrong implemented and tested improved ND coil geometry in swimmer, shown at right.

38. Swimmer Mods (Armstrong) No evidence of real improvement in tracking improvements. Changes have therefore not been added to development release.

39. Di-Muon Tracking Efficiency Current SR tracking code has not been optimized for multi-track finding. Improvements will require a better hough track finder implementation (volunteers welcome) Scanning of di-muon events provided by Costas shows that this this study will require every bit of performance we can squeeze out of the track finder. (and then some) However, some failures are unexplainable, and should be easily addressed.

40. Di-Muon Tracking Efficiencies

41. Alignment Far detector realignment completed for Release 1.18 –2 cm bug in 1% of strips fixed –2 cm error in starting width in all modules fixed (related to the module gap model) Vulcan data extensively studied by Louis Barrett & Phil Schreiner –Conclusion – cannot be used to improve alignment MC study by T. Fields indicates specification for alignment tolerances: 2-3 mm

42. Tracking Summary Working towards adoption of new fitter in 4-6 weeks. John will look into removing range constraints in fitter to remove these biases. Other areas that need checking/improvement –Dimuon reco –Effect of changes to fit chi2 calculation on CR muon analysis.

43. Misc. Known Issues X-talk removal. –Currently, x-talk is not removed prior to shower/track reconstruction. Mike has done some recent work on this. AltdeMux provides tools as well, as Mark T. reminds us. Andy Blake is interested in working on this, and we should study this for the next release. Address slicing errors in which small events are generated in the tail of larger ones.

44. Summary of Near Term Issues Validate new fitter (John) –Test removal of range constraint, understand failures, document. Restrict time window for showers in sub-shower code. (Chris) Fix ‘rock muon’ vertex problem (Jim, Mike) Implement x-talk removal Implement low pulse height removal. Improve di-muon tracking efficiency. Reduce frequency of ‘runt event’ formation in slicer. Others? Many problems are presented for the first time in parallel sessions that I may miss, so to the analysis group leaders – please tell me if I have missed something important that needs immediate attention.

45. Further Future As Jon mentioned yesterday, it would be useful at this point to step back and look at whether early assumptions for how we interact with the data are valid. This must be done in such a way that we end up with something that is truly better, not just different. This will require great care and time in the initial stages. However, my own view is that we should begin consideration of a major revision of the offline.